[Mesa-dev] [PATCH] r600/shader: when using images always load thread id gpr at start (v2)

[Gert Wollny]

For me this also fixed 

  tests/spec/arb_compute_shader/execution/
           multiple-texture-reading.shader_test

With that Tested-by: Gert Wollny <gw.fossdev at gmail.com>

Am Mittwoch, den 28.02.2018, 16:47 +1000 schrieb Dave Airlie:
> From: Dave Airlie <airlied at redhat.com>
> 
> The delayed loading code was fail if we had control flow.
> 
> This fixes:
> tests/spec/arb_shader_image_load_store/execution/image_checkerboard.s
> hader_test
> 
> v2: don't use temp_reg before setting temp_reg up.
> 
> Signed-off-by: Dave Airlie <airlied at redhat.com>
> ---
>  src/gallium/drivers/r600/r600_shader.c | 22 +++++++---------------
>  1 file changed, 7 insertions(+), 15 deletions(-)
> 
> diff --git a/src/gallium/drivers/r600/r600_shader.c
> b/src/gallium/drivers/r600/r600_shader.c
> index 1256e271b5b..5ddc5987326 100644
> --- a/src/gallium/drivers/r600/r600_shader.c
> +++ b/src/gallium/drivers/r600/r600_shader.c
> @@ -363,7 +363,6 @@ struct r600_shader_ctx {
>  	unsigned                                tess_input_info; /*
> temp with tess input offsets */
>  	unsigned                                tess_output_info; /*
> temp with tess input offsets */
>  	unsigned                                thread_id_gpr; /*
> temp with thread id calculated for images */
> -	bool thread_id_gpr_loaded;
>  };
>  
>  struct r600_shader_tgsi_instruction {
> @@ -3275,9 +3274,6 @@ static int load_thread_id_gpr(struct
> r600_shader_ctx *ctx)
>  	struct r600_bytecode_alu alu;
>  	int r;
>  
> -	if (ctx->thread_id_gpr_loaded)
> -		return 0;
> -
>  	memset(&alu, 0, sizeof(struct r600_bytecode_alu));
>  	alu.op = ALU_OP1_MBCNT_32LO_ACCUM_PREV_INT;
>  	alu.dst.sel = ctx->temp_reg;
> @@ -3322,7 +3318,6 @@ static int load_thread_id_gpr(struct
> r600_shader_ctx *ctx)
>  			   ctx->temp_reg, 0);
>  	if (r)
>  		return r;
> -	ctx->thread_id_gpr_loaded = true;
>  	return 0;
>  }
>  
> @@ -3431,12 +3426,12 @@ static int r600_shader_from_tgsi(struct
> r600_context *rctx,
>  	ctx.gs_next_vertex = 0;
>  	ctx.gs_stream_output_info = &so;
>  
> +	ctx.thread_id_gpr = -1;
>  	ctx.face_gpr = -1;
>  	ctx.fixed_pt_position_gpr = -1;
>  	ctx.fragcoord_input = -1;
>  	ctx.colors_used = 0;
>  	ctx.clip_vertex_write = 0;
> -	ctx.thread_id_gpr_loaded = false;
>  
>  	ctx.helper_invoc_reg = -1;
>  	ctx.cs_block_size_reg = -1;
> @@ -3570,7 +3565,6 @@ static int r600_shader_from_tgsi(struct
> r600_context *rctx,
>  
>  	if (shader->uses_images) {
>  		ctx.thread_id_gpr = ++regno;
> -		ctx.thread_id_gpr_loaded = false;
>  	}
>  	ctx.temp_reg = ++regno;
>  
> @@ -3613,6 +3607,12 @@ static int r600_shader_from_tgsi(struct
> r600_context *rctx,
>  	if (shader->vs_as_gs_a)
>  		vs_add_primid_output(&ctx, key.vs.prim_id_out);
>  
> +	if (ctx.thread_id_gpr != -1) {
> +		r = load_thread_id_gpr(&ctx);
> +		if (r)
> +			return r;
> +	}
> +
>  	if (ctx.type == PIPE_SHADER_TESS_EVAL)
>  		r600_fetch_tess_io_info(&ctx);
>  
> @@ -8647,10 +8647,6 @@ static int tgsi_load_rat(struct
> r600_shader_ctx *ctx)
>  	unsigned rat_index_mode;
>  	unsigned immed_base;
>  
> -	r = load_thread_id_gpr(ctx);
> -	if (r)
> -		return r;
> -
>  	rat_index_mode = inst->Src[0].Indirect.Index == 2 ? 2 : 0;
> // CF_INDEX_1 : CF_INDEX_NONE
>  
>  	immed_base = R600_IMAGE_IMMED_RESOURCE_OFFSET;
> @@ -8978,10 +8974,6 @@ static int tgsi_atomic_op_rat(struct
> r600_shader_ctx *ctx)
>  	immed_base = R600_IMAGE_IMMED_RESOURCE_OFFSET;
>  	rat_base = ctx->shader->rat_base;
>  
> -	r = load_thread_id_gpr(ctx);
> -	if (r)
> -		return r;
> -
>          if (inst->Src[0].Register.File == TGSI_FILE_BUFFER) {
>  		immed_base += ctx->info.file_count[TGSI_FILE_IMAGE];
>  		rat_base += ctx->info.file_count[TGSI_FILE_IMAGE];