[Mesa-dev] [PATCH v4 13/14] nvc0: expose ARB_compute_variable_group_size

[Ilia Mirkin]

On Wed, Oct 5, 2016 at 2:48 PM, Samuel Pitoiset
<samuel.pitoiset at gmail.com> wrote:
> Only expose 512 threads/block on Fermi to not be limited by
> 32 GPRs/thread.
>
> v4: - use 512 threads on Fermi, 2014 on Kepler+

Dyslexics... untie!

>
> Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
> ---
>  src/gallium/drivers/nouveau/nvc0/nvc0_screen.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c b/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
> index df6c6af..afcb08b 100644
> --- a/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
> +++ b/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
> @@ -448,6 +448,12 @@ nvc0_screen_get_compute_param(struct pipe_screen *pscreen,
>        RET(((uint64_t []) { 1024, 1024, 64 }));
>     case PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK:
>        RET((uint64_t []) { 1024 });
> +   case PIPE_COMPUTE_CAP_MAX_VARIABLE_THREADS_PER_BLOCK:
> +      if (obj_class >= NVE4_COMPUTE_CLASS) {
> +         RET((uint64_t []) { 1024 });
> +      } else {
> +         RET((uint64_t []) { 512 });
> +      }
>     case PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE: /* g[] */
>        RET((uint64_t []) { 1ULL << 40 });
>     case PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE: /* s[] */
> @@ -478,8 +484,6 @@ nvc0_screen_get_compute_param(struct pipe_screen *pscreen,
>        RET((uint32_t []) { 512 }); /* FIXME: arbitrary limit */
>     case PIPE_COMPUTE_CAP_ADDRESS_BITS:
>        RET((uint32_t []) { 64 });
> -   case PIPE_COMPUTE_CAP_MAX_VARIABLE_THREADS_PER_BLOCK:
> -      RET((uint64_t []) { 0 });
>     default:
>        return 0;
>     }
> --
> 2.10.0
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev