radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
/*
|
|
|
|
* Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
|
|
* to deal in the Software without restriction, including without limitation
|
|
|
|
* on the rights to use, copy, modify, merge, publish, distribute, sub
|
|
|
|
* license, and/or sell copies of the Software, and to permit persons to whom
|
|
|
|
* the Software is furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice (including the next
|
|
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
|
|
* Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
|
|
|
|
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
|
|
|
|
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
|
|
|
|
* USE OR OTHER DEALINGS IN THE SOFTWARE.
|
|
|
|
*/
|
2014-01-22 18:50:36 +01:00
|
|
|
|
|
|
|
#include "si_pipe.h"
|
|
|
|
#include "si_public.h"
|
2014-08-07 21:14:31 +02:00
|
|
|
#include "sid.h"
|
2014-01-22 18:50:36 +01:00
|
|
|
|
|
|
|
#include "radeon/radeon_uvd.h"
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
#include "util/u_memory.h"
|
|
|
|
#include "vl/vl_decoder.h"
|
|
|
|
|
|
|
|
/*
|
|
|
|
* pipe_context
|
|
|
|
*/
|
2014-01-07 03:07:55 +01:00
|
|
|
static void si_destroy_context(struct pipe_context *context)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
{
|
2014-01-11 16:00:50 +01:00
|
|
|
struct si_context *sctx = (struct si_context *)context;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
si_release_all_descriptors(sctx);
|
2013-08-06 06:42:22 +02:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
pipe_resource_reference(&sctx->null_const_buf.buffer, NULL);
|
|
|
|
r600_resource_reference(&sctx->border_color_table, NULL);
|
2012-09-07 16:09:08 +02:00
|
|
|
|
2014-03-08 14:18:30 +01:00
|
|
|
si_pm4_delete_state(sctx, gs_rings, sctx->gs_rings);
|
|
|
|
si_pm4_delete_state(sctx, gs_onoff, sctx->gs_on);
|
|
|
|
si_pm4_delete_state(sctx, gs_onoff, sctx->gs_off);
|
2013-11-21 16:45:28 +09:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
if (sctx->dummy_pixel_shader) {
|
|
|
|
sctx->b.b.delete_fs_state(&sctx->b.b, sctx->dummy_pixel_shader);
|
2012-08-08 15:35:42 +02:00
|
|
|
}
|
2014-08-23 03:25:29 +02:00
|
|
|
sctx->b.b.delete_depth_stencil_alpha_state(&sctx->b.b, sctx->custom_dsa_flush);
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->b.b.delete_blend_state(&sctx->b.b, sctx->custom_blend_resolve);
|
|
|
|
sctx->b.b.delete_blend_state(&sctx->b.b, sctx->custom_blend_decompress);
|
2014-03-06 02:44:07 +01:00
|
|
|
sctx->b.b.delete_blend_state(&sctx->b.b, sctx->custom_blend_fastclear);
|
2014-03-04 17:49:39 +01:00
|
|
|
util_unreference_framebuffer_state(&sctx->framebuffer.state);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
util_blitter_destroy(sctx->blitter);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-03-07 21:52:51 +01:00
|
|
|
si_pm4_cleanup(sctx);
|
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
r600_common_context_cleanup(&sctx->b);
|
|
|
|
FREE(sctx);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|
|
|
|
|
2014-01-07 03:07:55 +01:00
|
|
|
static struct pipe_context *si_create_context(struct pipe_screen *screen, void *priv)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
{
|
2014-01-11 16:00:50 +01:00
|
|
|
struct si_context *sctx = CALLOC_STRUCT(si_context);
|
2014-01-11 16:01:11 +01:00
|
|
|
struct si_screen* sscreen = (struct si_screen *)screen;
|
2014-04-11 22:14:27 +02:00
|
|
|
struct radeon_winsys *ws = sscreen->b.ws;
|
2013-10-30 20:44:23 +01:00
|
|
|
int shader, i;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
if (sctx == NULL)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
return NULL;
|
|
|
|
|
2014-01-21 18:01:01 +01:00
|
|
|
sctx->b.b.screen = screen; /* this must be set first */
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->b.b.priv = priv;
|
|
|
|
sctx->b.b.destroy = si_destroy_context;
|
2014-01-21 18:01:01 +01:00
|
|
|
sctx->screen = sscreen; /* Easy accessing of screen/winsys. */
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-21 18:01:01 +01:00
|
|
|
if (!r600_common_context_init(&sctx->b, &sscreen->b))
|
|
|
|
goto fail;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
si_init_blit_functions(sctx);
|
|
|
|
si_init_compute_functions(sctx);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-11 16:01:11 +01:00
|
|
|
if (sscreen->b.info.has_uvd) {
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->b.b.create_video_codec = si_uvd_create_decoder;
|
|
|
|
sctx->b.b.create_video_buffer = si_video_buffer_create;
|
2013-04-03 10:18:35 +02:00
|
|
|
} else {
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->b.b.create_video_codec = vl_create_decoder;
|
|
|
|
sctx->b.b.create_video_buffer = vl_video_buffer_create;
|
2013-04-03 10:18:35 +02:00
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-04-12 17:34:44 +02:00
|
|
|
sctx->b.rings.gfx.cs = ws->cs_create(ws, RING_GFX, si_context_gfx_flush,
|
2014-04-11 22:14:27 +02:00
|
|
|
sctx, NULL);
|
2014-04-12 17:34:44 +02:00
|
|
|
sctx->b.rings.gfx.flush = si_context_gfx_flush;
|
2013-08-06 06:42:22 +02:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
si_init_all_descriptors(sctx);
|
2013-08-06 06:42:22 +02:00
|
|
|
|
radeonsi: simplify and improve flushing
This mimics r600g. The R600_CONTEXT_xxx flags are added to rctx->b.flags
and si_emit_cache_flush emits the packets. That's it. The shared radeon code
tells us when the streamout cache should be flushed, so we have to check
the flags anyway.
There is a new atom "cache_flush", because caches must be flushed *after*
resource descriptors are changed in memory.
Functional changes:
* Write caches are flushed at the end of CS and read caches are flushed
at its beginning.
* Sampler view states are removed from si_state, they only held the flush
flags.
* Everytime a shader is changed, the I cache is flushed. Is this needed?
Due to a hw bug, this also flushes the K cache.
* The WRITE_DATA packet is changed to use TC, which fixes a rendering issue
in openarena. I'm not sure how TC interacts with CP DMA, but for now it
seems to work better than any other solution I tried. (BTW CIK allows us
to use TC for CP DMA.)
* Flush the K cache instead of the texture cache when updating resource
descriptors (due to a hw bug, this also flushes the I cache).
I think the K cache flush is correct here, but I'm not sure if the texture
cache should be flushed too (probably not considering we use TC
for WRITE_DATA, but we don't use TC for CP DMA).
* The number of resource contexts is decreased to 16. With all of these cache
changes, 4 doesn't work, but 8 works, which suggests I'm actually doing
the right thing here and the pipeline isn't drained during flushes.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>
2013-08-26 17:19:39 +02:00
|
|
|
/* Initialize cache_flush. */
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->cache_flush = si_atom_cache_flush;
|
2014-04-22 12:46:08 -04:00
|
|
|
sctx->atoms.s.cache_flush = &sctx->cache_flush;
|
radeonsi: simplify and improve flushing
This mimics r600g. The R600_CONTEXT_xxx flags are added to rctx->b.flags
and si_emit_cache_flush emits the packets. That's it. The shared radeon code
tells us when the streamout cache should be flushed, so we have to check
the flags anyway.
There is a new atom "cache_flush", because caches must be flushed *after*
resource descriptors are changed in memory.
Functional changes:
* Write caches are flushed at the end of CS and read caches are flushed
at its beginning.
* Sampler view states are removed from si_state, they only held the flush
flags.
* Everytime a shader is changed, the I cache is flushed. Is this needed?
Due to a hw bug, this also flushes the K cache.
* The WRITE_DATA packet is changed to use TC, which fixes a rendering issue
in openarena. I'm not sure how TC interacts with CP DMA, but for now it
seems to work better than any other solution I tried. (BTW CIK allows us
to use TC for CP DMA.)
* Flush the K cache instead of the texture cache when updating resource
descriptors (due to a hw bug, this also flushes the I cache).
I think the K cache flush is correct here, but I'm not sure if the texture
cache should be flushed too (probably not considering we use TC
for WRITE_DATA, but we don't use TC for CP DMA).
* The number of resource contexts is decreased to 16. With all of these cache
changes, 4 doesn't work, but 8 works, which suggests I'm actually doing
the right thing here and the pipeline isn't drained during flushes.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>
2013-08-26 17:19:39 +02:00
|
|
|
|
2014-05-06 13:48:11 +02:00
|
|
|
sctx->msaa_config = si_atom_msaa_config;
|
|
|
|
sctx->atoms.s.msaa_config = &sctx->msaa_config;
|
|
|
|
|
2014-04-22 12:46:08 -04:00
|
|
|
sctx->atoms.s.streamout_begin = &sctx->b.streamout.begin_atom;
|
|
|
|
sctx->atoms.s.streamout_enable = &sctx->b.streamout.enable_atom;
|
2013-08-14 01:52:38 +02:00
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
switch (sctx->b.chip_class) {
|
2013-06-07 13:58:34 -04:00
|
|
|
case SI:
|
2013-06-07 14:00:11 -04:00
|
|
|
case CIK:
|
2014-01-11 16:00:50 +01:00
|
|
|
si_init_state_functions(sctx);
|
|
|
|
si_init_config(sctx);
|
2013-06-07 14:00:11 -04:00
|
|
|
break;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
default:
|
2014-01-11 16:00:50 +01:00
|
|
|
R600_ERR("Unsupported chip class %d.\n", sctx->b.chip_class);
|
2013-08-14 01:04:39 +02:00
|
|
|
goto fail;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->blitter = util_blitter_create(&sctx->b.b);
|
|
|
|
if (sctx->blitter == NULL)
|
2013-08-14 01:04:39 +02:00
|
|
|
goto fail;
|
2014-08-18 00:55:40 +02:00
|
|
|
sctx->blitter->draw_rectangle = r600_draw_rectangle;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2013-08-31 00:13:43 +02:00
|
|
|
/* these must be last */
|
2014-01-11 16:00:50 +01:00
|
|
|
si_begin_new_cs(sctx);
|
2014-01-22 00:06:32 +01:00
|
|
|
r600_query_init_backend_mask(&sctx->b); /* this emits commands and must be last */
|
2013-10-30 20:44:23 +01:00
|
|
|
|
|
|
|
/* CIK cannot unbind a constant buffer (S_BUFFER_LOAD is buggy
|
|
|
|
* with a NULL buffer). We need to use a dummy buffer instead. */
|
2014-01-11 16:00:50 +01:00
|
|
|
if (sctx->b.chip_class == CIK) {
|
|
|
|
sctx->null_const_buf.buffer = pipe_buffer_create(screen, PIPE_BIND_CONSTANT_BUFFER,
|
2014-02-03 03:42:17 +01:00
|
|
|
PIPE_USAGE_DEFAULT, 16);
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->null_const_buf.buffer_size = sctx->null_const_buf.buffer->width0;
|
2013-10-30 20:44:23 +01:00
|
|
|
|
|
|
|
for (shader = 0; shader < SI_NUM_SHADERS; shader++) {
|
2014-07-07 23:27:19 +02:00
|
|
|
for (i = 0; i < SI_NUM_CONST_BUFFERS; i++) {
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->b.b.set_constant_buffer(&sctx->b.b, shader, i,
|
|
|
|
&sctx->null_const_buf);
|
2013-10-30 20:44:23 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Clear the NULL constant buffer, because loads should return zeros. */
|
2014-01-11 16:00:50 +01:00
|
|
|
sctx->b.clear_buffer(&sctx->b.b, sctx->null_const_buf.buffer, 0,
|
|
|
|
sctx->null_const_buf.buffer->width0, 0);
|
2013-10-30 20:44:23 +01:00
|
|
|
}
|
|
|
|
|
2014-01-11 16:00:50 +01:00
|
|
|
return &sctx->b.b;
|
2013-08-14 01:04:39 +02:00
|
|
|
fail:
|
2014-01-11 16:00:50 +01:00
|
|
|
si_destroy_context(&sctx->b.b);
|
2013-08-14 01:04:39 +02:00
|
|
|
return NULL;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* pipe_screen
|
|
|
|
*/
|
|
|
|
|
2014-01-07 03:07:55 +01:00
|
|
|
static int si_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
{
|
2014-01-11 16:01:11 +01:00
|
|
|
struct si_screen *sscreen = (struct si_screen *)pscreen;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
|
|
|
switch (param) {
|
|
|
|
/* Supported features (boolean caps). */
|
|
|
|
case PIPE_CAP_TWO_SIDED_STENCIL:
|
2012-04-14 17:37:37 -04:00
|
|
|
case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS:
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_ANISOTROPIC_FILTER:
|
|
|
|
case PIPE_CAP_POINT_SPRITE:
|
|
|
|
case PIPE_CAP_OCCLUSION_QUERY:
|
|
|
|
case PIPE_CAP_TEXTURE_SHADOW_MAP:
|
|
|
|
case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
|
|
|
|
case PIPE_CAP_BLEND_EQUATION_SEPARATE:
|
|
|
|
case PIPE_CAP_TEXTURE_SWIZZLE:
|
|
|
|
case PIPE_CAP_DEPTH_CLIP_DISABLE:
|
|
|
|
case PIPE_CAP_SHADER_STENCIL_EXPORT:
|
|
|
|
case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR:
|
|
|
|
case PIPE_CAP_MIXED_COLORBUFFER_FORMATS:
|
|
|
|
case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT:
|
|
|
|
case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
|
|
|
|
case PIPE_CAP_SM3:
|
|
|
|
case PIPE_CAP_SEAMLESS_CUBE_MAP:
|
|
|
|
case PIPE_CAP_PRIMITIVE_RESTART:
|
|
|
|
case PIPE_CAP_CONDITIONAL_RENDER:
|
|
|
|
case PIPE_CAP_TEXTURE_BARRIER:
|
|
|
|
case PIPE_CAP_INDEP_BLEND_ENABLE:
|
|
|
|
case PIPE_CAP_INDEP_BLEND_FUNC:
|
|
|
|
case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE:
|
|
|
|
case PIPE_CAP_VERTEX_COLOR_UNCLAMPED:
|
2012-04-24 01:23:33 +02:00
|
|
|
case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY:
|
|
|
|
case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY:
|
|
|
|
case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY:
|
2012-04-24 15:19:31 +02:00
|
|
|
case PIPE_CAP_USER_INDEX_BUFFERS:
|
|
|
|
case PIPE_CAP_USER_CONSTANT_BUFFERS:
|
2012-06-18 22:50:02 +02:00
|
|
|
case PIPE_CAP_START_INSTANCE:
|
2012-09-05 18:27:02 +02:00
|
|
|
case PIPE_CAP_NPOT_TEXTURES:
|
2013-10-04 04:32:15 -04:00
|
|
|
case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES:
|
2013-03-14 17:18:43 +01:00
|
|
|
case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER:
|
2013-03-21 18:02:52 +01:00
|
|
|
case PIPE_CAP_TGSI_INSTANCEID:
|
2012-10-25 13:50:10 -04:00
|
|
|
case PIPE_CAP_COMPUTE:
|
2013-10-31 15:08:49 +01:00
|
|
|
case PIPE_CAP_TEXTURE_BUFFER_OBJECTS:
|
2014-07-03 11:15:18 -04:00
|
|
|
case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT:
|
2014-01-22 00:08:11 +01:00
|
|
|
case PIPE_CAP_QUERY_PIPELINE_STATISTICS:
|
2014-01-27 21:46:21 +01:00
|
|
|
case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT:
|
2014-04-22 21:23:29 +02:00
|
|
|
case PIPE_CAP_CUBE_MAP_ARRAY:
|
2014-05-07 13:15:41 +02:00
|
|
|
case PIPE_CAP_SAMPLE_SHADING:
|
2014-04-24 03:03:43 +02:00
|
|
|
case PIPE_CAP_DRAW_INDIRECT:
|
2013-08-17 02:47:21 +02:00
|
|
|
return 1;
|
2013-08-16 15:21:45 +02:00
|
|
|
|
2013-07-30 22:29:30 +02:00
|
|
|
case PIPE_CAP_TEXTURE_MULTISAMPLE:
|
2013-11-20 13:48:19 +01:00
|
|
|
/* 2D tiling on CIK is supported since DRM 2.35.0 */
|
2014-07-08 02:50:57 +02:00
|
|
|
return sscreen->b.chip_class < CIK ||
|
|
|
|
sscreen->b.info.drm_minor >= 35;
|
2013-07-30 22:29:30 +02:00
|
|
|
|
2012-10-28 17:52:48 +01:00
|
|
|
case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT:
|
2014-03-09 22:29:20 +01:00
|
|
|
return R600_MAP_BUFFER_ALIGNMENT;
|
2012-10-28 17:52:48 +01:00
|
|
|
|
2012-04-24 17:31:17 +02:00
|
|
|
case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT:
|
2014-03-09 20:05:54 +01:00
|
|
|
case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT:
|
|
|
|
return 4;
|
2012-04-24 17:31:17 +02:00
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_GLSL_FEATURE_LEVEL:
|
2014-07-08 02:50:57 +02:00
|
|
|
return 330;
|
2013-10-31 15:08:49 +01:00
|
|
|
|
|
|
|
case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE:
|
2014-01-11 16:01:11 +01:00
|
|
|
return MIN2(sscreen->b.info.vram_size, 0xFFFFFFFF);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-06-16 16:53:42 +02:00
|
|
|
case PIPE_CAP_TEXTURE_QUERY_LOD:
|
2014-06-06 03:00:18 +02:00
|
|
|
case PIPE_CAP_TEXTURE_GATHER_SM5:
|
|
|
|
return HAVE_LLVM >= 0x0305;
|
|
|
|
case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS:
|
|
|
|
return HAVE_LLVM >= 0x0305 ? 4 : 0;
|
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
/* Unsupported features. */
|
|
|
|
case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT:
|
|
|
|
case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER:
|
|
|
|
case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS:
|
|
|
|
case PIPE_CAP_FRAGMENT_COLOR_CLAMPED:
|
|
|
|
case PIPE_CAP_VERTEX_COLOR_CLAMPED:
|
|
|
|
case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION:
|
2012-04-24 01:23:33 +02:00
|
|
|
case PIPE_CAP_USER_VERTEX_BUFFERS:
|
2014-03-09 22:29:20 +01:00
|
|
|
case PIPE_CAP_TGSI_TEXCOORD:
|
2013-11-27 19:47:51 +10:00
|
|
|
case PIPE_CAP_FAKE_SW_MSAA:
|
2014-04-26 23:44:57 -04:00
|
|
|
case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
|
2014-05-17 01:20:19 +02:00
|
|
|
case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
|
2014-08-14 00:04:41 -04:00
|
|
|
case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
|
2014-08-17 03:37:19 +02:00
|
|
|
case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
|
2014-08-20 19:45:10 -04:00
|
|
|
case PIPE_CAP_SAMPLER_VIEW_TARGET:
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
return 0;
|
|
|
|
|
2013-10-30 21:44:07 +01:00
|
|
|
case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK:
|
|
|
|
return PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_R600;
|
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
/* Stream output. */
|
|
|
|
case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS:
|
2014-01-11 16:01:11 +01:00
|
|
|
return sscreen->b.has_streamout ? 4 : 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME:
|
2014-01-11 16:01:11 +01:00
|
|
|
return sscreen->b.has_streamout ? 1 : 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS:
|
|
|
|
case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS:
|
2014-01-11 16:01:11 +01:00
|
|
|
return sscreen->b.has_streamout ? 32*4 : 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-02-09 22:56:20 +01:00
|
|
|
/* Geometry shader output. */
|
|
|
|
case PIPE_CAP_MAX_GEOMETRY_OUTPUT_VERTICES:
|
|
|
|
return 1024;
|
|
|
|
case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS:
|
|
|
|
return 4095;
|
2014-06-26 20:01:50 -04:00
|
|
|
case PIPE_CAP_MAX_VERTEX_STREAMS:
|
|
|
|
return 1;
|
2014-02-09 22:56:20 +01:00
|
|
|
|
2014-08-19 21:09:58 -10:00
|
|
|
case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE:
|
|
|
|
return 2048;
|
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
/* Texturing. */
|
|
|
|
case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
|
|
|
|
case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
|
2014-03-09 20:03:57 +01:00
|
|
|
return 15; /* 16384 */
|
|
|
|
case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
|
|
|
|
/* textures support 8192, but layered rendering supports 2048 */
|
|
|
|
return 12;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS:
|
2014-03-09 20:03:57 +01:00
|
|
|
/* textures support 8192, but layered rendering supports 2048 */
|
|
|
|
return 2048;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
|
|
|
/* Render targets. */
|
|
|
|
case PIPE_CAP_MAX_RENDER_TARGETS:
|
|
|
|
return 8;
|
|
|
|
|
2013-09-21 19:45:08 +02:00
|
|
|
case PIPE_CAP_MAX_VIEWPORTS:
|
|
|
|
return 1;
|
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
/* Timer queries, present when the clock frequency is non zero. */
|
2013-08-09 11:59:28 +02:00
|
|
|
case PIPE_CAP_QUERY_TIMESTAMP:
|
2012-12-09 09:50:34 +00:00
|
|
|
case PIPE_CAP_QUERY_TIME_ELAPSED:
|
2014-01-11 16:01:11 +01:00
|
|
|
return sscreen->b.info.r600_clock_crystal_freq != 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-04-09 14:58:53 -04:00
|
|
|
case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET:
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_MIN_TEXEL_OFFSET:
|
2014-06-06 03:00:18 +02:00
|
|
|
return -32;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-04-09 14:58:53 -04:00
|
|
|
case PIPE_CAP_MAX_TEXTURE_GATHER_OFFSET:
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_CAP_MAX_TEXEL_OFFSET:
|
2014-06-06 03:00:18 +02:00
|
|
|
return 31;
|
|
|
|
|
2013-07-09 21:21:39 -07:00
|
|
|
case PIPE_CAP_ENDIANNESS:
|
|
|
|
return PIPE_ENDIAN_LITTLE;
|
2014-08-14 20:57:29 +01:00
|
|
|
|
|
|
|
case PIPE_CAP_VENDOR_ID:
|
|
|
|
return 0x1002;
|
|
|
|
case PIPE_CAP_DEVICE_ID:
|
|
|
|
return sscreen->b.info.pci_id;
|
|
|
|
case PIPE_CAP_ACCELERATED:
|
|
|
|
return 1;
|
|
|
|
case PIPE_CAP_VIDEO_MEMORY:
|
|
|
|
return sscreen->b.info.vram_size >> 20;
|
|
|
|
case PIPE_CAP_UMA:
|
|
|
|
return 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-01-07 03:07:55 +01:00
|
|
|
static int si_get_shader_param(struct pipe_screen* pscreen, unsigned shader, enum pipe_shader_cap param)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
{
|
|
|
|
switch(shader)
|
|
|
|
{
|
|
|
|
case PIPE_SHADER_FRAGMENT:
|
|
|
|
case PIPE_SHADER_VERTEX:
|
|
|
|
case PIPE_SHADER_GEOMETRY:
|
2014-01-24 17:51:34 +09:00
|
|
|
break;
|
2012-10-25 13:50:10 -04:00
|
|
|
case PIPE_SHADER_COMPUTE:
|
|
|
|
switch (param) {
|
|
|
|
case PIPE_SHADER_CAP_PREFERRED_IR:
|
|
|
|
return PIPE_SHADER_IR_LLVM;
|
2014-06-17 08:52:34 -07:00
|
|
|
case PIPE_SHADER_CAP_DOUBLES:
|
|
|
|
return 0; /* XXX: Enable doubles once the compiler can
|
|
|
|
handle them. */
|
2014-08-07 15:31:17 -04:00
|
|
|
case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE: {
|
|
|
|
uint64_t max_const_buffer_size;
|
|
|
|
pscreen->get_compute_param(pscreen,
|
|
|
|
PIPE_COMPUTE_CAP_MAX_MEM_ALLOC_SIZE,
|
|
|
|
&max_const_buffer_size);
|
|
|
|
return max_const_buffer_size;
|
|
|
|
}
|
2012-10-25 13:50:10 -04:00
|
|
|
default:
|
|
|
|
return 0;
|
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
default:
|
|
|
|
/* TODO: support tessellation */
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (param) {
|
|
|
|
case PIPE_SHADER_CAP_MAX_INSTRUCTIONS:
|
|
|
|
case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS:
|
|
|
|
case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS:
|
|
|
|
case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS:
|
|
|
|
return 16384;
|
|
|
|
case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH:
|
2013-01-31 19:40:24 +01:00
|
|
|
return 32;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_MAX_INPUTS:
|
2014-07-07 23:27:19 +02:00
|
|
|
return shader == PIPE_SHADER_VERTEX ? SI_NUM_VERTEX_BUFFERS : 32;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_MAX_TEMPS:
|
|
|
|
return 256; /* Max native temporaries. */
|
2014-07-24 20:32:08 +02:00
|
|
|
case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE:
|
|
|
|
return 4096 * sizeof(float[4]); /* actually only memory limits this */
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_MAX_CONST_BUFFERS:
|
2014-07-07 23:27:19 +02:00
|
|
|
return SI_NUM_USER_CONST_BUFFERS;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_MAX_PREDS:
|
|
|
|
return 0; /* FIXME */
|
|
|
|
case PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED:
|
|
|
|
return 1;
|
2013-02-01 11:16:54 -07:00
|
|
|
case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED:
|
|
|
|
return 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR:
|
2014-01-22 18:47:21 +09:00
|
|
|
/* Indirection of geometry shader input dimension is not
|
|
|
|
* handled yet
|
|
|
|
*/
|
|
|
|
return shader < PIPE_SHADER_GEOMETRY;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR:
|
|
|
|
case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR:
|
|
|
|
case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR:
|
2013-03-07 12:00:18 +01:00
|
|
|
return 1;
|
2012-07-25 08:22:30 -04:00
|
|
|
case PIPE_SHADER_CAP_INTEGERS:
|
|
|
|
return 1;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
case PIPE_SHADER_CAP_SUBROUTINES:
|
|
|
|
return 0;
|
|
|
|
case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS:
|
2013-11-26 02:30:41 +01:00
|
|
|
case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS:
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
return 16;
|
2012-10-11 21:35:45 +02:00
|
|
|
case PIPE_SHADER_CAP_PREFERRED_IR:
|
|
|
|
return PIPE_SHADER_IR_TGSI;
|
2014-06-17 08:52:34 -07:00
|
|
|
case PIPE_SHADER_CAP_DOUBLES:
|
|
|
|
return 0;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-01-07 03:07:55 +01:00
|
|
|
static void si_destroy_screen(struct pipe_screen* pscreen)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
{
|
2014-01-11 16:01:11 +01:00
|
|
|
struct si_screen *sscreen = (struct si_screen *)pscreen;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
|
2014-01-11 16:01:11 +01:00
|
|
|
if (sscreen == NULL)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
return;
|
|
|
|
|
2014-04-09 00:26:32 +02:00
|
|
|
if (!sscreen->b.ws->unref(sscreen->b.ws))
|
2013-09-25 13:59:56 +02:00
|
|
|
return;
|
|
|
|
|
2014-01-22 02:49:53 +01:00
|
|
|
r600_destroy_common_screen(&sscreen->b);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|
|
|
|
|
2014-08-07 21:14:31 +02:00
|
|
|
#define SI_TILE_MODE_COLOR_2D_8BPP 14
|
|
|
|
|
|
|
|
/* Initialize pipe config. This is especially important for GPUs
|
|
|
|
* with 16 pipes and more where it's initialized incorrectly by
|
|
|
|
* the TILING_CONFIG ioctl. */
|
|
|
|
static bool si_initialize_pipe_config(struct si_screen *sscreen)
|
|
|
|
{
|
|
|
|
unsigned mode2d;
|
|
|
|
|
|
|
|
/* This is okay, because there can be no 2D tiling without
|
|
|
|
* the tile mode array, so we won't need the pipe config.
|
|
|
|
* Return "success".
|
|
|
|
*/
|
|
|
|
if (!sscreen->b.info.si_tile_mode_array_valid)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
/* The same index is used for the 2D mode on CIK too. */
|
|
|
|
mode2d = sscreen->b.info.si_tile_mode_array[SI_TILE_MODE_COLOR_2D_8BPP];
|
|
|
|
|
|
|
|
switch (G_009910_PIPE_CONFIG(mode2d)) {
|
|
|
|
case V_02803C_ADDR_SURF_P2:
|
|
|
|
sscreen->b.tiling_info.num_channels = 2;
|
|
|
|
break;
|
|
|
|
case V_02803C_X_ADDR_SURF_P4_8X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P4_16X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P4_16X32:
|
|
|
|
case V_02803C_X_ADDR_SURF_P4_32X32:
|
|
|
|
sscreen->b.tiling_info.num_channels = 4;
|
|
|
|
break;
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_16X16_8X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_16X32_8X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_32X32_8X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_16X32_16X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_32X32_16X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_32X32_16X32:
|
|
|
|
case V_02803C_X_ADDR_SURF_P8_32X64_32X32:
|
|
|
|
sscreen->b.tiling_info.num_channels = 8;
|
|
|
|
break;
|
|
|
|
case V_02803C_X_ADDR_SURF_P16_32X32_8X16:
|
|
|
|
case V_02803C_X_ADDR_SURF_P16_32X32_16X16:
|
|
|
|
sscreen->b.tiling_info.num_channels = 16;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
assert(0);
|
|
|
|
fprintf(stderr, "radeonsi: Unknown pipe config %i.\n",
|
|
|
|
G_009910_PIPE_CONFIG(mode2d));
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
struct pipe_screen *radeonsi_screen_create(struct radeon_winsys *ws)
|
|
|
|
{
|
2014-01-11 16:01:11 +01:00
|
|
|
struct si_screen *sscreen = CALLOC_STRUCT(si_screen);
|
|
|
|
if (sscreen == NULL) {
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2013-09-22 21:47:35 +02:00
|
|
|
/* Set functions first. */
|
2014-01-11 16:01:11 +01:00
|
|
|
sscreen->b.b.context_create = si_create_context;
|
|
|
|
sscreen->b.b.destroy = si_destroy_screen;
|
|
|
|
sscreen->b.b.get_param = si_get_param;
|
|
|
|
sscreen->b.b.get_shader_param = si_get_shader_param;
|
|
|
|
sscreen->b.b.is_format_supported = si_is_format_supported;
|
2014-01-28 06:51:50 -08:00
|
|
|
sscreen->b.b.resource_create = r600_resource_create_common;
|
2013-09-22 21:47:35 +02:00
|
|
|
|
2014-08-07 21:14:31 +02:00
|
|
|
if (!r600_common_screen_init(&sscreen->b, ws) ||
|
|
|
|
!si_initialize_pipe_config(sscreen)) {
|
2014-01-11 16:01:11 +01:00
|
|
|
FREE(sscreen);
|
2013-09-22 22:12:18 +02:00
|
|
|
return NULL;
|
|
|
|
}
|
2013-09-22 21:47:35 +02:00
|
|
|
|
2014-01-11 16:01:11 +01:00
|
|
|
sscreen->b.has_cp_dma = true;
|
2014-07-08 02:50:57 +02:00
|
|
|
sscreen->b.has_streamout = true;
|
2013-11-28 15:09:35 +01:00
|
|
|
|
2013-09-22 21:47:35 +02:00
|
|
|
if (debug_get_bool_option("RADEON_DUMP_SHADERS", FALSE))
|
2014-01-11 16:01:11 +01:00
|
|
|
sscreen->b.debug_flags |= DBG_FS | DBG_VS | DBG_GS | DBG_PS | DBG_CS;
|
2013-09-22 21:47:35 +02:00
|
|
|
|
2013-10-03 16:39:50 +02:00
|
|
|
/* Create the auxiliary context. This must be done last. */
|
2014-01-11 16:01:11 +01:00
|
|
|
sscreen->b.aux_context = sscreen->b.b.context_create(&sscreen->b.b, NULL);
|
2013-10-03 16:39:50 +02:00
|
|
|
|
2014-01-11 16:01:11 +01:00
|
|
|
return &sscreen->b.b;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 17:38:37 -05:00
|
|
|
}
|