[Mesa-dev] [PATCH 14/14] radeonsi/gfx9: allow the scratch buffer in HS and GS

[Nicolai Hähnle]

On 28.04.2017 23:42, Marek Olšák wrote:
> From: Marek Olšák <marek.olsak at amd.com>
>
> It works now.

Nice! Patches 8-14 are also:

Reviewed-by: Nicolai Hähnle <nicolai.haehnle at amd.com>


> ---
>  src/gallium/drivers/radeonsi/si_state_shaders.c | 10 ----------
>  1 file changed, 10 deletions(-)
>
> diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
> index a3ff511..a5260f5 100644
> --- a/src/gallium/drivers/radeonsi/si_state_shaders.c
> +++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
> @@ -491,25 +491,20 @@ static void si_shader_hs(struct si_screen *sscreen, struct si_shader *shader)
>  	if (sscreen->b.chip_class >= GFX9) {
>  		si_pm4_set_reg(pm4, R_00B410_SPI_SHADER_PGM_LO_LS, va >> 8);
>  		si_pm4_set_reg(pm4, R_00B414_SPI_SHADER_PGM_HI_LS, va >> 40);
>
>  		/* We need at least 2 components for LS.
>  		 * VGPR0-3: (VertexID, RelAutoindex, InstanceID / StepRate0, InstanceID).
>  		 * StepRate0 is set to 1. so that VGPR3 doesn't have to be loaded.
>  		 */
>  		ls_vgpr_comp_cnt = shader->info.uses_instanceid ? 2 : 1;
>
> -		if (shader->config.scratch_bytes_per_wave) {
> -			fprintf(stderr, "HS: scratch buffer unsupported");
> -			abort();
> -		}
> -
>  		shader->config.rsrc2 =
>  			S_00B42C_USER_SGPR(GFX9_TCS_NUM_USER_SGPR) |
>  			S_00B42C_USER_SGPR_MSB(GFX9_TCS_NUM_USER_SGPR >> 5) |
>  			S_00B42C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0);
>  	} else {
>  		si_pm4_set_reg(pm4, R_00B420_SPI_SHADER_PGM_LO_HS, va >> 8);
>  		si_pm4_set_reg(pm4, R_00B424_SPI_SHADER_PGM_HI_HS, va >> 40);
>
>  		shader->config.rsrc2 =
>  			S_00B42C_USER_SGPR(GFX6_TCS_NUM_USER_SGPR) |
> @@ -809,25 +804,20 @@ static void si_shader_gs(struct si_screen *sscreen, struct si_shader *shader)
>  		si_pm4_set_reg(pm4, R_028A94_VGT_GS_MAX_PRIMS_PER_SUBGROUP,
>  			       S_028A94_MAX_PRIMS_PER_SUBGROUP(gs_info.max_prims_per_subgroup));
>  		si_pm4_set_reg(pm4, R_028AAC_VGT_ESGS_RING_ITEMSIZE,
>  			       shader->key.part.gs.es->esgs_itemsize / 4);
>
>  		if (es_type == PIPE_SHADER_TESS_EVAL)
>  			si_set_tesseval_regs(sscreen, shader->key.part.gs.es, pm4);
>
>  		polaris_set_vgt_vertex_reuse(sscreen, shader->key.part.gs.es,
>  					     NULL, pm4);
> -
> -		if (shader->config.scratch_bytes_per_wave) {
> -			fprintf(stderr, "GS: scratch buffer unsupported");
> -			abort();
> -		}
>  	} else {
>  		si_pm4_set_reg(pm4, R_00B220_SPI_SHADER_PGM_LO_GS, va >> 8);
>  		si_pm4_set_reg(pm4, R_00B224_SPI_SHADER_PGM_HI_GS, va >> 40);
>
>  		si_pm4_set_reg(pm4, R_00B228_SPI_SHADER_PGM_RSRC1_GS,
>  			       S_00B228_VGPRS((shader->config.num_vgprs - 1) / 4) |
>  			       S_00B228_SGPRS((shader->config.num_sgprs - 1) / 8) |
>  			       S_00B228_DX10_CLAMP(1) |
>  			       S_00B228_FLOAT_MODE(shader->config.float_mode));
>  		si_pm4_set_reg(pm4, R_00B22C_SPI_SHADER_PGM_RSRC2_GS,
>


-- 
Lerne, wie die Welt wirklich ist,
Aber vergiss niemals, wie sie sein sollte.