This is the only sane solution for nv50 and nvc0 (really, trust me),
but since on other hardware the border colour is tightly coupled with
texture state they'd have to undo the swizzle, so I've added a cap.
The dependency of update_sampler on the texture updates was
introduced to avoid doing the apply_depthmode to the swizzle twice.
v2: Moved swizzling helper to u_format.c, extended the CAP to
provide more accurate information.
TGSI_OPCODE_IF condition had two possible interpretations:
- src.x != 0.0f
- Mesa statetracker when PIPE_SHADER_CAP_INTEGERS was false either for
vertex and fragment shaders
- gallivm/llvmpipe
- postprocess
- vl state tracker
- vega state tracker
- most old drivers
- old internal state trackers
- many graw examples
- src.x != 0U
- Mesa statetracker when PIPE_SHADER_CAP_INTEGERS was true for both
vertex and fragment shaders
- tgsi_exec/softpipe
- r600
- radeonsi
- nv50
And drivers that use draw module also were a mess (because Mesa would
emit float IFs, but draw module supports native integers so it would
interpret IF arg as integers...)
This sort of works if the source argument is limited to float +0.0f or
+1.0f, integer 0, but would fail if source is float -0.0f, or integer in
the float NaN range. It could also fail if source is integer 1, and
hardware flushes denormalized numbers to zero.
But with this change there are now two opcodes, IF and UIF, with clear
meaning.
Drivers that do not support native integers do not need to worry about
UIF. However, for backwards compatibility with old state trackers and
examples, it is advisable that native integer capable drivers also
support the float IF opcode.
I tried to implement this for r600 and radeonsi based on the surrounding
code. I couldn't do this for nouveau, so I just shunted IF/UIF
together, which matches the current behavior.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
v2:
- Incorporate Roland's feedback.
- Fix r600_shader.c merge conflict.
- Fix typo in radeon, spotted by Michel Dänzer.
- Incorporte Christoph Bumiller's patch to handle TGSI_OPCODE_IF(float)
properly in nv50/ir.
This target string now contains four values instead of three. The old
processor field (which was really being interpreted as arch) has been split
into two fields: processor and arch. This allows drivers to pass a
more a more detailed description of the hardware to compiler frontends.
v2:
- Adapt to libclc changes
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
The blit-based paths for TexImage, GetTexImage, and ReadPixels aren't very
fast with software rasterizer. Now Gallium drivers have the ability to turn
them off.
Reviewed-by: Brian Paul <brianp@vmware.com>
Tested-by: Brian Paul <brianp@vmware.com>
This makes it possible to identify gl_TexCoord and gl_PointCoord
for drivers where sprite coordinate replacement is restricted.
The new PIPE_CAP_TGSI_TEXCOORD decides whether these varyings
should be hidden behind the GENERIC semantic or not.
With this patch only nvc0 and nv30 will request that they be used.
v2: introduce a CAP so other drivers don't have to bother with
the new semantic
v3: adapt to introduction gl_varying_slot enum
This clarifies some things and gets rid of some old stuff.
The most significant one is probably that buffers cannot have formats
(nearly all drivers completely ignored format and used width0 as byte size
already in any case). There seems to be no use case for "structured" buffers.
(Note while d3d11 has new Structured Buffers, these still aren't associated
with a format, rather a byte stride, which we can't do yet either way.)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Need to take the type into account. Also, if we want to allow
mov's with modifiers we need to pick a type (assume float).
v2: don't allow all modifiers on all type, in particular don't allow
absolute on non-float types and don't allow negate on unsigned.
Also treat UADD as signed (despite the name) since it is used
for handling both signed and unsigned integer arguments and otherwise
modifiers don't work.
Also add tgsi docs clarifying this.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
It looks like using coord.w as explicit lod value is a mistake, most likely
because some dx10 docs had it specified that way. Seems this was changed though:
http://msdn.microsoft.com/en-us/library/windows/desktop/hh447229%28v=vs.85%29.aspx
- let's just hope it doesn't depend on runtime build version or something.
Not only would this need translation (so go against the stated goal these
opcodes should be close to dx10 semantics) but it would prevent usage of this
opcode with cube arrays, which is apparently possible:
http://msdn.microsoft.com/en-us/library/windows/desktop/bb509699%28v=vs.85%29.aspx
(Note not only does this show cube arrays using explicit lod, but also the
confusion with this opcode: it lists an explicit lod parameter value, but then
states last component of location is used as lod).
(For "true" hw drivers, only nv50 had code to handle it, and it appears the
code was already right for the new semantics, though fix up the seemingly
wrong c/d arguments while there.)
v2: fix comment, separate out other changes.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Need to calculate the number of mip levels (if it would be worthwile could
store it in dynamic state).
While here, the query code also used chan 2 for the lod value.
This worked with mesa state tracker but it seems safer to use chan 0.
Still passes piglit textureSize (with some handwaving), though the non-GL
parts are (largely) untested.
v2: clarify and expect the sviewinfo opcode to return ints, not floats,
just like the OpenGL textureSize (dx10 supports dst modifiers with resinfo).
Also simplify some code.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
The glsl-to-tgsi translater will emit SQRT to implement GLSL's sqrt()
and distance() functions if the PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
query says it's supported by the driver.
Otherwise, sqrt(x) is implemented with x*rsq(x). The problem with
this is sqrt(0) must be handled specially because rsq(0) might be
Inf/NaN/undefined (and then 0*rsq(0) is Inf/Nan/undefined). In the
glsl-to-tgsi code we use an extra CMP to check if x is zero and then
replace the result of x*rsq(x) with zero.
In the end, this makes sqrt() generate much more reasonable code for
drivers that can do square roots.
Note that many of piglit's generated shader tests use the GLSL
distance() function.
Reviewed-by: Brian Paul <brianp@vmware.com>
v2: update relnotes-9.1
v3: use align_malloc and align_free for malloced buffers in r300g
v4: document the new CAP in the docs
"get_transfer + transfer_map" becomes "transfer_map".
"transfer_unmap + transfer_destroy" becomes "transfer_unmap".
transfer_map must create and return the transfer object and transfer_unmap
must destroy it.
transfer_map is successful if the returned buffer pointer is not NULL.
If transfer_map fails, the pointer to the transfer object remains unchanged
(i.e. doesn't have to be NULL).
Acked-by: Brian Paul <brianp@vmware.com>
Used to implement proxy textures. If a gallium driver doesn't implement
this function we'll just continue to use the core Mesa fallback code.
Without this hook we really have no good way to implement OpenGL proxy
textures with gallium drivers.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
v2: Tom Stellard
- Update CAP description
v3: Tom Stellard
- TGSI targets should pass an empty string for this CAP.
v4: Tom Stellard
- TGSI targets can ignore this CAP.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Add a shader cap for specifying the preferred shader representation.
Right now the only supported value is TGSI, other enum values will be
added as they are needed.
This is mainly to accommodate AMD's LLVM compiler back-end by letting
it bypass the TGSI representation for compute programs. Other drivers
will keep using the common TGSI instruction set.
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
This change will be useful to implement function parameter passing on
top of TGSI. As we don't have a proper stack, a register-based
calling convention will be used instead, which isn't necessarily a bad
thing given that GPUs often have plenty of registers to spare.
Using the same register space for local temporaries and
inter-procedural communication caused some inefficiencies, because in
some cases the register allocator would lose the freedom to merge
temporary values together into the same physical register, leading to
suboptimal register (and sometimes, as a side effect, instruction)
usage.
The LOCAL declaration modifier specifies that the value isn't intended
for parameter passing and as a result the compiler doesn't have to
give any guarantees of it being preserved across function boundaries.
Ignoring the LOCAL flag doesn't change the semantics of a valid
program in any way, because local variables are just supposed to get a
more relaxed treatment. IOW, this should be a backwards-compatible
change.
Normal resource access (e.g. the LOAD TGSI opcode) is supposed to
perform a series of conversions to turn the texture data as it's found
in memory into the target data type.
In compute programs it's often the case that we only want to access
the raw bits as they're stored in some buffer object, and any kind of
channel conversion and scaling is harmful or inefficient, especially
in implementations that lack proper hardware support to take care of
it -- in those cases the conversion has to be implemented in software
and it's likely to result in a performance hit even if the pipe_buffer
and declaration data types are set up in a way that would just pass
the data through.
Add a declaration flag that marks a resource as typeless. No channel
conversion will be performed in that case, and the X coordinate of the
address vector will be interpreted in byte units instead of elements
for obvious reasons.
This is similar to D3D11's ByteAddressBuffer, and will be used to
implement OpenCL's constant arguments. The remaining four compute
memory spaces can also be understood as raw resources.
Move Interpolate, Centroid and CylindricalWrap from tgsi_declaration
to a separate token -- they only make sense for FS inputs and we need
room for other flags in the top-level declaration token.
This commit splits the current concept of resource into "sampler
views" and "shader resources":
"Sampler views" are textures or buffers that are bound to a given
shader stage and can be read from in conjunction with a sampler
object. They are analogous to OpenGL texture objects or Direct3D
SRVs.
"Shader resources" are textures or buffers that can be read and
written from a shader. There's no support for floating point
coordinates, address wrap modes or filtering, and, unlike sampler
views, shader resources are global for the whole graphics pipeline.
They are analogous to OpenGL image objects (as in
ARB_shader_image_load_store) or Direct3D UAVs.
Most hardware is likely to implement shader resources and sampler
views as separate objects, so, having the distinction at the API level
simplifies things slightly for the driver.
This patch introduces the SVIEW register file with a declaration token
and syntax analogous to the already existing RES register file. After
this change, the SAMPLE_* opcodes no longer accept a resource as
input, but rather a SVIEW object. To preserve the functionality of
reading from a sampler view with integer coordinates, the
SAMPLE_I(_MS) opcodes are introduced which are similar to LOAD(_MS)
but take a SVIEW register instead of a RES register as argument.
Define an interface that exposes the minimal functionality required to
implement some of the popular compute APIs. This commit adds entry
points to set the grid layout and other state required to keep track
of the usual address spaces employed in compute APIs, to bind a
compute program, and execute it on the device.
Reviewed-by: Marek Olšák <maraeo@gmail.com>
This is required for any serious constant buffer support.
Constant buffer offsets on ATI and NVIDIA DX10 and DX11 GPUs must be
a multiple of 256.
In OpenGL, this can be queried via GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT.
