[Mesa-dev] [PATCH 00/11] add support for ARB_compute_variable_group_size

[Samuel Pitoiset]

On 09/09/2016 09:43 AM, Nicolai Hähnle wrote:
> Hi Samuel,

Hi,

>
> thanks for working on this! I've only skimmed the series so far without
> looking at each patch in detail:
>
> For GCN, we need to pass the (maximum) work group size (aka block size /
> local group size) into the compiler, because it affects occupancy and
> hence register limits. And the lowering of TGSI_SEMANTIC_BLOCK_SIZE also
> needs to be changed. This means more work will definitely be needed for
> radeonsi.

I will add a new cap for MaxComputeVariableGroupInvocations as suggested 
by Marek. And enable the extension based on that cap.

That way, you will be able to fix up radeonsi in a separate patch.

>
> It would be great if you could add docs for
> TGSI_SEMANTIC_{GROUP,BLOCK}_SIZE to tgsi.rst. I know they were already
> missing before.

That would be great yeah, but in a separate patch out of that series I 
think.

>
> Also, the nouveau enablement patch should come _before_ the patch that
> turns on the extension...

Good catch, thanks. :)

>
> Cheers,
> Nicolai
>
> On 08.09.2016 22:31, Samuel Pitoiset wrote:
>> Hi,
>>
>> This series implements ARB_compute_variable_group_size written against
>> GL 4.3.
>> This extension allows to dispatch variable work group size via a new
>> function
>> called glDispatchComputeGroupSizeARB().
>>
>> Because this extension is pretty similar to ARB_compute_shader, all
>> Gallium
>> drivers which already support compute shaders will expose
>> ARB_compute_variable_group_size with that series.
>>
>> I did write a bunch of piglit tests, have a look here if you want:
>> https://lists.freedesktop.org/archives/piglit/2016-September/020755.html
>>
>> All tests pass on Fermi (GF119) as well as all previous compute
>> shaders tests.
>>
>> Marek, Nicolai and other AMD folks, I don't know if radeonsi will need
>> a fix
>> somewhere for handling a variable work group size, but as I don't have
>> the
>> hardware, I can't test. Let me know if something needs to be slighty
>> updated.
>>
>> Please review,
>> Thanks!
>>
>> Samuel Pitoiset (11):
>>   glapi: add entry points for GL_ARB_compute_variable_group_size
>>   mesa/main: add support for ARB_compute_variable_groups_size
>>   glsl: add enable flags for ARB_compute_variable_group_size
>>   glsl: process local_size_variable input qualifier
>>   glsl: reject compute shaders with fixed and variable local size
>>   glsl/linker: handle errors when a variable local size is used
>>   glsl: add gl_LocalGroupSizeARB as a system value
>>   st/mesa: add mapping for SYSTEM_VALUE_LOCAL_GROUP_SIZE
>>   st/mesa: add support for dispatching a variable local size
>>   st/mesa: expose ARB_compute_variable_group_size
>>   nv50/ir: use 1024 threads/block for variable local size
>>
>>  src/compiler/glsl/ast.h                            |  5 ++
>>  src/compiler/glsl/ast_to_hir.cpp                   | 14 ++++
>>  src/compiler/glsl/ast_type.cpp                     |  6 ++
>>  src/compiler/glsl/builtin_variables.cpp            |  2 +
>>  src/compiler/glsl/glsl_parser.yy                   | 13 +++
>>  src/compiler/glsl/glsl_parser_extras.cpp           |  6 ++
>>  src/compiler/glsl/glsl_parser_extras.h             |  8 ++
>>  src/compiler/glsl/linker.cpp                       | 23 +++++-
>>  src/compiler/glsl/standalone.cpp                   |  4 +
>>  src/compiler/glsl/standalone_scaffolding.cpp       |  5 ++
>>  src/compiler/shader_enums.h                        |  1 +
>>  .../drivers/nouveau/codegen/nv50_ir_target.h       |  3 +-
>>  .../glapi/gen/ARB_compute_variable_group_size.xml  | 25 ++++++
>>  src/mapi/glapi/gen/Makefile.am                     |  1 +
>>  src/mapi/glapi/gen/gl_API.xml                      |  2 +
>>  src/mesa/main/api_validate.c                       | 94
>> ++++++++++++++++++++++
>>  src/mesa/main/api_validate.h                       |  4 +
>>  src/mesa/main/compute.c                            | 25 ++++++
>>  src/mesa/main/compute.h                            |  5 ++
>>  src/mesa/main/context.c                            |  6 ++
>>  src/mesa/main/dd.h                                 |  9 +++
>>  src/mesa/main/extensions_table.h                   |  1 +
>>  src/mesa/main/get.c                                | 12 +++
>>  src/mesa/main/get_hash_params.py                   |  3 +
>>  src/mesa/main/mtypes.h                             | 23 +++++-
>>  src/mesa/main/shaderapi.c                          |  1 +
>>  src/mesa/main/shaderobj.c                          |  2 +
>>  src/mesa/main/tests/dispatch_sanity.cpp            |  3 +
>>  src/mesa/state_tracker/st_cb_compute.c             | 15 +++-
>>  src/mesa/state_tracker/st_extensions.c             | 13 +++
>>  src/mesa/state_tracker/st_glsl_to_tgsi.cpp         |  2 +
>>  31 files changed, 329 insertions(+), 7 deletions(-)
>>  create mode 100644
>> src/mapi/glapi/gen/ARB_compute_variable_group_size.xml
>>