[Mesa-dev] [PATCH 3/3] tgsi/scan: set correct input limits for geometry shader

Roland Scheidegger sroland at vmware.com
Mon Apr 22 15:42:10 PDT 2013


Am 20.04.2013 09:04, schrieb Zack Rusin:
> TGSI geometry shader input declerations are of the IN[][2] format
> and the dimensions of the array have to be deduced from the input
> primitive property.
> 
> Signed-off-by: Zack Rusin <zackr at vmware.com>
> ---
>  src/gallium/auxiliary/tgsi/tgsi_scan.c |   17 +++++++++++++++++
>  1 file changed, 17 insertions(+)
> 
> diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c
> index 373391d..bd79405 100644
> --- a/src/gallium/auxiliary/tgsi/tgsi_scan.c
> +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
> @@ -36,6 +36,7 @@
>  
>  #include "util/u_debug.h"
>  #include "util/u_math.h"
> +#include "util/u_prim.h"
>  #include "tgsi/tgsi_parse.h"
>  #include "tgsi/tgsi_util.h"
>  #include "tgsi/tgsi_scan.h"
> @@ -261,6 +262,22 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
>        case TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS:
>           info->color0_writes_all_cbufs = info->properties[i].data[0];
>           break;
> +      case TGSI_PROPERTY_GS_INPUT_PRIM:
> +         /* The dimensions of the IN decleration in geometry shader have
> +          * to be deduced from the type of the input primitive.
> +          */
> +         if (procType == TGSI_PROCESSOR_GEOMETRY) {
> +            unsigned input_primitive = info->properties[i].data[0];
> +            int num_verts = u_vertices_per_prim(input_primitive);
> +            unsigned j;
> +            info->file_count[TGSI_FILE_INPUT] = num_verts;
> +            info->file_max[TGSI_FILE_INPUT] =
> +               MAX2(info->file_max[TGSI_FILE_INPUT], num_verts - 1);
> +            for (j = 0; j < num_verts; ++j) {
> +               info->file_mask[TGSI_FILE_INPUT] |= (1 << j);
> +            }
> +         }
> +         break;
>        default:
>           ;
>        }
> 

Can you even have that property in in a non-geometry shader?
In any case, series looks good to me, though your patches seem to appear
with some retro-date.

Roland


More information about the mesa-dev mailing list