[Mesa-dev] [PATCH] i965/vec4: Use the sampler for pull constant loads on Broadwell.

[Jordan Justen]

Reviewed-by: Jordan Justen <jordan.l.justen at intel.com>

On Sat, Jun 14, 2014 at 12:58 PM, Kenneth Graunke <kenneth at whitecape.org> wrote:
> We've used the LD sampler message for pull constant loads on earlier
> hardware for some time, and also were already using it for the FS on
> Broadwell.  This patch makes us use it for Broadwell VS/GS as well.
>
> I believe that when I wrote this code in 2012, we still used the data
> port in some cases, and I somehow neglected to convert it while
> rebasing.
>
> Improves performance in GLBenchmark 2.7 Egypt by 416.978% +/- 2.25821%
> (n = 17).  Many other applications should benefit similarly: this speeds
> up uniform array access in the VS, which is commonly used for skinning
> shaders, among other things.
>
> Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> ---
>  src/mesa/drivers/dri/i965/gen8_vec4_generator.cpp | 16 ++++++++--------
>  1 file changed, 8 insertions(+), 8 deletions(-)
>
> No Piglit regressions.
>
> diff --git a/src/mesa/drivers/dri/i965/gen8_vec4_generator.cpp b/src/mesa/drivers/dri/i965/gen8_vec4_generator.cpp
> index 14070cd..82ea45a 100644
> --- a/src/mesa/drivers/dri/i965/gen8_vec4_generator.cpp
> +++ b/src/mesa/drivers/dri/i965/gen8_vec4_generator.cpp
> @@ -444,14 +444,14 @@ gen8_vec4_generator::generate_pull_constant_load(vec4_instruction *inst,
>     gen8_instruction *send = next_inst(BRW_OPCODE_SEND);
>     gen8_set_dst(brw, send, dst);
>     gen8_set_src0(brw, send, offset);
> -   gen8_set_dp_message(brw, send, GEN7_SFID_DATAPORT_DATA_CACHE,
> -                       surf_index,
> -                       GEN6_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ,
> -                       0,      /* message control */
> -                       1,      /* mlen */
> -                       1,      /* rlen */
> -                       false,  /* no header */
> -                       false); /* EOT */
> +   gen8_set_sampler_message(brw, send,
> +                            surf_index,
> +                            0, /* The LD message ignores the sampler unit. */
> +                            GEN5_SAMPLER_MESSAGE_SAMPLE_LD,
> +                            1,      /* rlen */
> +                            1,      /* mlen */
> +                            false,  /* no header */
> +                            BRW_SAMPLER_SIMD_MODE_SIMD4X2);
>
>     brw_mark_surface_used(&prog_data->base, surf_index);
>  }
> --
> 1.9.1
>