[Mesa-dev] [PATCH 3/6] i965/fs: Properly report regs_written from SAMPLEINFO

[Jason Ekstrand]

On Apr 23, 2016 4:39 PM, "Kenneth Graunke" <kenneth at whitecape.org> wrote:
>
> From: Jason Ekstrand <jason.ekstrand at intel.com>
>
> Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
> ---
>  src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 8 +++++++-
>  src/mesa/drivers/dri/i965/brw_shader.cpp | 3 ++-
>  2 files changed, 9 insertions(+), 2 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> index 725f5da..43d3745 100644
> --- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
> @@ -3205,12 +3205,18 @@ fs_visitor::nir_emit_texture(const fs_builder
&bld, nir_tex_instr *instr)
>     case nir_texop_txs: op = ir_txs; break;
>     case nir_texop_texture_samples: {
>        fs_reg dst = retype(get_nir_dest(instr->dest),
BRW_REGISTER_TYPE_D);
> -      fs_inst *inst = bld.emit(SHADER_OPCODE_SAMPLEINFO, dst,
> +
> +      fs_reg tmp = bld.vgrf(BRW_REGISTER_TYPE_D, 4);

This one should Cc stable.  The previous behavior would only allocate one
register and then write four thus potentially stomping three innocent
bystanders.  I've never seen an actual bug related to it but sampleinfo is
newish.

> +      fs_inst *inst = bld.emit(SHADER_OPCODE_SAMPLEINFO, tmp,
>                                 bld.vgrf(BRW_REGISTER_TYPE_D, 1),
>                                 texture_reg, texture_reg);
>        inst->mlen = 1;
>        inst->header_size = 1;
>        inst->base_mrf = -1;
> +      inst->regs_written = 4 * (dispatch_width / 8);
> +
> +      /* Pick off the one component we care about */
> +      bld.MOV(dst, tmp);
>        return;
>     }
>     case nir_texop_samples_identical: op = ir_samples_identical; break;
> diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp
b/src/mesa/drivers/dri/i965/brw_shader.cpp
> index 376cb25..76c4a52 100644
> --- a/src/mesa/drivers/dri/i965/brw_shader.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
> @@ -710,7 +710,8 @@ backend_instruction::is_tex() const
>             opcode == SHADER_OPCODE_TXS ||
>             opcode == SHADER_OPCODE_LOD ||
>             opcode == SHADER_OPCODE_TG4 ||
> -           opcode == SHADER_OPCODE_TG4_OFFSET);
> +           opcode == SHADER_OPCODE_TG4_OFFSET ||
> +           opcode == SHADER_OPCODE_SAMPLEINFO);
>  }
>
>  bool
> --
> 2.8.0
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160423/4c46b281/attachment-0001.html>