Commit Graph

33810 Commits

Author SHA1 Message Date
Eric Anholt
64516430be i965: Fix scissoring when width or height is 0.
We would run into trouble due to the hardware using inclusive numbers
and the subtraction to handle that producing negative (meaning large
positive) coordinates.

Bug #27643.
2010-04-22 15:37:33 -07:00
Zack Rusin
021e0dc78b gallivm: implement indirect addressing over temporaries
a bit more involved than indirect addressing over consts, but still
fairly reasonable. we allocate an array instead of individual alloca's,
and we do it only if the shader does indirect addressing.
2010-04-22 18:36:07 -04:00
Zack Rusin
ded2374e67 gallivm: implement indirect addressing over constants
implement indirect addressing (ARL and ARR instructions) when used
with CONST's. indirect addressing over other vars (temps, inputs, outputs)
is not supported yet.
2010-04-22 17:33:14 -04:00
Tormod Volden
355370e30b radeon: 9800 SE has only one quadpipe
Although these cards have 2 pipelines on the silicon only
the first passed the QA and the other should be disabled.

http://www.digital-daily.com/video/ati-radeon9800se/
http://www.rojakpot.com/showarticle.aspx?artno=101&pgno=1

Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
2010-04-22 17:14:15 -04:00
Brian Paul
d888812324 gallium: use better param name 2010-04-22 14:39:12 -06:00
Brian Paul
19b26cb0a2 gallivm: remove Z/stencil special case code in lp_build_sample_offset()
Fixes progs/tests/zreaddraw.c and progs/demos/shadowtex.c
2010-04-22 14:37:50 -06:00
Brian Paul
17249ae8e0 st/mesa: assert that binding flags are properly set for drawing surfaces 2010-04-22 14:37:50 -06:00
Brian Paul
e9bcad4c33 st/mesa: fix binding flags in st_ChooseTextureFormat()
See comment for more info.
2010-04-22 14:37:50 -06:00
Brian Paul
9f74fa111b progs/tests: asst. improvements to zreaddraw.c 2010-04-22 14:37:50 -06:00
Jesse Barnes
45e2b51c85 DRI2/GLX: check for vblank_mode in DRI2 GLX code
Re-add support for the vblank_mode environment and configuration
variable.  Useful for benchmarking and app control.
2010-04-22 12:49:45 -07:00
Jesse Barnes
234286c0f8 DRI2: add config query extension
Add a new DRI2 configuration query extension.  Allows for DRI2 client
code to query for common DRI2 configuration options.
2010-04-22 12:49:45 -07:00
José Fonseca
c059565054 llvmpipe: Undo zs tiling changes.
tile_w/tile_h have different meaning there.

This partialy reverts commit ccdc6b5913.
2010-04-22 20:38:36 +01:00
José Fonseca
cb384ae8b2 gallium/docs: Unswap PIPE_BIND_SHARED's comment. 2010-04-22 19:51:10 +01:00
Brian Paul
f836a5f60b gallium/docs: many updates to screen.rst
Put PIPE_BIND_ and PIPE_TEXTURE_GEOM_ prefixes on token names so
that they can be found with grep.  This needs to be done in more places.

Corrected/improved a lot of information and grammer.

I don't know how to properly format everything - someone else can take
care of that.
2010-04-22 12:47:10 -06:00
José Fonseca
cdaceaafee llvmpipe: Portability fixes. 2010-04-22 19:25:00 +01:00
José Fonseca
ccdc6b5913 llvmpipe: Fix rendering to non 32bpp formats.
Tiled layout always used 32bpp, therefore linear/tiled strides are not
related.
2010-04-22 19:25:00 +01:00
Zack Rusin
18a4a83dda gallivm: update comments 2010-04-22 13:57:18 -04:00
Zack Rusin
6d2e09fdc2 gallivm: fix nested cont statements 2010-04-22 13:57:18 -04:00
Zack Rusin
3a423dcf9d gallivm: fix nested break statemants 2010-04-22 13:57:18 -04:00
Zack Rusin
63aa9b505a gallivm: make sure we return the correct type when approximating log's 2010-04-22 13:57:18 -04:00
José Fonseca
8bee4c7718 llvmpipe: Use resource_is_texture() consistently.
Otherwise slightly difference order causes assertion failures.

Also remove mentions of PIPE_BIND_SCANOUT/PIPE_BIND_SHARED. They are not
propoer bind flags and will likely be deprecated. If surfaces should
be passed to the winsys then they should have the DISPLAY_TARGET flag
set, which is a proper bind flag.
2010-04-22 18:42:03 +01:00
Brian Paul
50d77c79fb gallium: fix comment language, remove dangling comment 2010-04-22 11:33:26 -06:00
Brian Paul
1143221cfd st/mesa: consolidate code for finding supported formats 2010-04-22 11:11:26 -06:00
Brian Paul
e65b11525a st/mesa: refactor depth/stencil format selection code 2010-04-22 11:11:26 -06:00
Brian Paul
446062b834 mesa: sort texel fetch/store table by format index 2010-04-22 11:10:42 -06:00
Brian Paul
b4a191e351 Merge branch '7.8'
Conflicts:

	src/mesa/state_tracker/st_format.c
2010-04-22 11:10:19 -06:00
Brian Paul
b1a9c76b57 st/mesa: add cases for MESA_FORMAT_Z24_X8, MESA_FORMAT_X8_Z24 2010-04-22 11:06:19 -06:00
Brian Paul
4d7ef6e06b mesa: fix conversion errors in signed_rgba8888[rev] texel fetch
Without the cast the returned texel colors were wrong.
Also, we don't need the "& 0xff" part anymore.
Bug found by Vinson Lee.
2010-04-22 11:06:19 -06:00
José Fonseca
2eea1714fd llvmpipe: Fix resource_is_texture.
It was missing PIPE_BIND_RENDER_TARGET, causing assertion failures for
pure render targets.

Also bind flags are too variable and complex for a good assessment for
whether the resource is a texture or not. Target is more concise.
2010-04-22 18:06:05 +01:00
José Fonseca
510669cd87 llvmpipe: Do not advertise S3TC rendering support. 2010-04-22 15:12:30 +01:00
José Fonseca
6ae9975ea0 llvmpipe: Fix typo in assert. 2010-04-22 15:11:57 +01:00
José Fonseca
52bc90caa8 u_format: Reduce code duplication. 2010-04-22 15:11:34 +01:00
Vinson Lee
bd91f665a7 gallivm: Remove unused variable. 2010-04-22 00:45:27 -07:00
Alex Deucher
48a1806126 r300g: fix vertex unit setup
RV3xx is 2, RV560,RV570 is 8

Noticed by Tormod Volden.
2010-04-22 02:37:34 -04:00
Alex Deucher
986eb4b99f r300: fix vertex unit setup
RV3xx is 2, RV560,RV570 is 8

Noticed by Tormod Volden.
2010-04-22 02:33:10 -04:00
Vinson Lee
7a0b48e349 gallium/util: Fix deviation calculation in u_cache.
The approximate deviation calculation was using the entire cache's total
instead of each cache entry's total.
2010-04-21 18:45:44 -07:00
Vinson Lee
393009a939 gallium/tests: Add test case for u_cache. 2010-04-21 18:31:01 -07:00
Brian Paul
7fe93f831d gallivm: added some assertions in loop-gen code
We're hitting these assertions with nested loops...
2010-04-21 18:22:18 -06:00
Aaron Plattner
2e995cdf25 progs/demos: Fix the progs/demos/rain help text
Signed-off-by: Brian Paul <brianp@vmware.com>
2010-04-21 18:13:19 -06:00
Brian Paul
bba6a196bb gallivm: fix copy&paste error: s/cont_stack_size/break_stack_size/ 2010-04-21 17:58:32 -06:00
Brian Paul
fc9b8cd9dd gallivm: emit_instruction() is boolean 2010-04-21 17:57:44 -06:00
Brian Paul
5b639cdc3e progs/glsl: better comments 2010-04-21 16:30:13 -06:00
Brian Paul
04c136d5c2 llvmpipe: add missing dependency in Makefile 2010-04-21 16:30:13 -06:00
Brian Paul
feffd259da gallivm: implement TGSI KILP
As in tgsi_exec.c we don't actually rely on condition codes; we do
an unconditional kill.  The only predication comes from the execution
mask which applies inside loops/conditionals.
2010-04-21 16:30:13 -06:00
José Fonseca
5c364b3ef7 draw: Make LLVM integration work without llvmpipe too (e.g. sofptipe).
Prevents the assertion failure.
2010-04-21 22:58:41 +01:00
Marek Olšák
dab8f6b1d9 radeon: fix warning 2010-04-21 22:57:01 +02:00
Marek Olšák
dd029dcb14 r300g: raise the max index for generic varyings to 32
This should fix an assertion failure in the game Heroes of Newearth.
Yes, the game seems to generate semantic indices greater than 15.
2010-04-21 22:57:01 +02:00
Brian Paul
01c12ed60c softpipe: call draw_wide_point_sprites() 2010-04-21 14:25:28 -06:00
Brian Paul
8cb223eb02 gallium/draw: fix point sprite handling
New draw API function to indicate whether or not to convert points to
quads for sprite rasterization.

Fix point-to-quad conversion regression in the wide-point stage.  We
need to check the pipe_rasterizer_state::point_quad_rasterization flag.
2010-04-21 14:25:28 -06:00
Brian Paul
d30ab4394e gallium/draw: use local var to simplify code 2010-04-21 14:25:28 -06:00