[Mesa-dev] [PATCH 1/3] gallium: add PIPE_CAP_SHAREABLE_SHADERS

[Marek Olšák]

On Tue, Oct 20, 2015 at 1:41 AM, Rob Clark <robdclark at gmail.com> wrote:
> On Mon, Oct 19, 2015 at 7:07 PM, Marek Olšák <maraeo at gmail.com> wrote:
>> On Mon, Oct 19, 2015 at 11:52 PM, Rob Clark <robdclark at gmail.com> wrote:
>>> On Sat, Oct 10, 2015 at 9:09 PM, Marek Olšák <maraeo at gmail.com> wrote:
>>>> From: Marek Olšák <marek.olsak at amd.com>
>>>>
>>>> I'll let drivers figure out how to do it.
>>>> ---
>>>>  src/gallium/docs/source/screen.rst               | 2 ++
>>>>  src/gallium/drivers/freedreno/freedreno_screen.c | 1 +
>>>>  src/gallium/drivers/i915/i915_screen.c           | 1 +
>>>>  src/gallium/drivers/ilo/ilo_screen.c             | 1 +
>>>>  src/gallium/drivers/llvmpipe/lp_screen.c         | 1 +
>>>>  src/gallium/drivers/nouveau/nv30/nv30_screen.c   | 1 +
>>>>  src/gallium/drivers/nouveau/nv50/nv50_screen.c   | 1 +
>>>>  src/gallium/drivers/nouveau/nvc0/nvc0_screen.c   | 1 +
>>>>  src/gallium/drivers/r300/r300_screen.c           | 1 +
>>>>  src/gallium/drivers/r600/r600_pipe.c             | 1 +
>>>>  src/gallium/drivers/radeonsi/si_pipe.c           | 1 +
>>>>  src/gallium/drivers/softpipe/sp_screen.c         | 1 +
>>>>  src/gallium/drivers/svga/svga_screen.c           | 1 +
>>>>  src/gallium/drivers/vc4/vc4_screen.c             | 1 +
>>>>  src/gallium/include/pipe/p_defines.h             | 1 +
>>>>  15 files changed, 16 insertions(+)
>>>>
>>>> diff --git a/src/gallium/docs/source/screen.rst b/src/gallium/docs/source/screen.rst
>>>> index e08844b..72f7596 100644
>>>> --- a/src/gallium/docs/source/screen.rst
>>>> +++ b/src/gallium/docs/source/screen.rst
>>>> @@ -276,6 +276,8 @@ The integer capabilities:
>>>>    GL4 hardware will likely need to emulate it with a shader variant, or by
>>>>    selecting the interpolation weights with a conditional assignment
>>>>    in the shader.
>>>> +* ``PIPE_CAP_SHAREABLE_SHADERS``: Whether shader CSOs can be used by any
>>>> +  pipe_context.
>>>
>>> An extra sentence along the lines that drivers supporting this cap can
>>> save pipe->screen ptr but not pipe ptr in the CSO might not hurt..
>>>
>>> other than that nitpick, lgtm..
>>
>> I don't understand why drivers would want to put either of them into
>> the CSO. All calls manipulating the CSO take pipe_context, so you
>> always have some pipe_context and pipe_screen if you need them. The
>> correct answer - don't put any of them into the CSO - applied before
>> and still applies.
>
> My point was, pipe_screen never changes for the CSO but pipe_context
> can..  I need the pipe_screen to stay the same for allocating memory
> to hold the shader, figure out which gpu generation for some things
> that differ between a3xx/a4xx in the backend, etc.
>
> Sure technically you could pass the screen down the call stack instead
> of storing it in CSO.. for me that would be a lot of adding extra fxn
> param's in a bunch of places, so I won't bother (but I would replace
> the current pipe_context ptr w/ pipe_screen, and update the emit code
> to pass down the ctx)

That's a fair point. Well, you can use whatever structure you want as
long as it's thread-safe. Even pipe_screen might not be thread-safe in
some drivers. :)

Marek