svga: Add support for arb_texture_gather

With sm4_1, we can support single channel 2D or CubeMap textures.
This patch exercises this feature.

Tested with piglit

v2: As per Brian's comment

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>

diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c
index b66b560fcc0bfca4935fb1d4f41bbf96430f1d63..a38e062f77683727e9d2b3c6389506362939876b 100644 (file)
--- a/src/gallium/drivers/svga/svga_screen.c
+++ b/src/gallium/drivers/svga/svga_screen.c
@@ -349,6 +349,9 @@ svga_get_param(struct pipe_screen *screen, enum pipe_cap param)
    case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
       return sws->have_sm4_1;
 
+   case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS:
+      return sws->have_sm4_1 ? 1 : 0; /* only single-channel textures */
+
    /* Unsupported features */
    case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
    case PIPE_CAP_TEXTURE_MIRROR_CLAMP_TO_EDGE:
@@ -362,7 +365,6 @@ svga_get_param(struct pipe_screen *screen, enum pipe_cap param)
    case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT:
    case PIPE_CAP_QUERY_PIPELINE_STATISTICS:
    case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT:
-   case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS:
    case PIPE_CAP_TEXTURE_GATHER_SM5:
    case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT:
    case PIPE_CAP_TEXTURE_QUERY_LOD:

diff --git a/src/gallium/drivers/svga/svga_tgsi_vgpu10.c b/src/gallium/drivers/svga/svga_tgsi_vgpu10.c
index fa27edcd7cf1da0aeddbe3d5d5ef108dda529a1f..b87ac9065843f024083128a71148e40953af8876 100644 (file)
--- a/src/gallium/drivers/svga/svga_tgsi_vgpu10.c
+++ b/src/gallium/drivers/svga/svga_tgsi_vgpu10.c
@@ -5294,6 +5294,52 @@ emit_tex(struct svga_shader_emitter_v10 *emit,
    return TRUE;
 }
 
+/**
+ * Emit code for TGSI_OPCODE_TG4 (texture lookup for texture gather)
+ */
+static boolean
+emit_tg4(struct svga_shader_emitter_v10 *emit,
+         const struct tgsi_full_instruction *inst)
+{
+   const uint unit = inst->Src[2].Register.Index;
+   unsigned target = inst->Texture.Texture;
+   struct tgsi_full_src_register coord;
+   int offsets[3];
+   struct tex_swizzle_info swz_info;
+
+   /* check that the sampler returns a float */
+   if (!is_valid_tex_instruction(emit, inst))
+      return TRUE;
+
+   if (target == TGSI_TEXTURE_CUBE_ARRAY) {
+      debug_printf("TGSI_TEXTURE_CUBE_ARRAY is not supported\n");
+      return TRUE;
+   }
+
+   begin_tex_swizzle(emit, unit, inst, FALSE, &swz_info);
+
+   get_texel_offsets(emit, inst, offsets);
+
+   coord = setup_texcoord(emit, unit, &inst->Src[0]);
+
+   /* Gather dst, coord, resource, sampler */
+   begin_emit_instruction(emit);
+   emit_sample_opcode(emit, VGPU10_OPCODE_GATHER4,
+                      inst->Instruction.Saturate, offsets);
+   emit_dst_register(emit, get_tex_swizzle_dst(&swz_info));
+   emit_src_register(emit, &coord);
+   emit_resource_register(emit, unit);
+   emit_sampler_register(emit, unit);
+   end_emit_instruction(emit);
+
+   end_tex_swizzle(emit, &swz_info);
+
+   free_temp_indexes(emit);
+
+   return TRUE;
+}
+
+
 
 /**
  * Emit code for TGSI_OPCODE_TEX2 (texture lookup for shadow cube map arrays)
@@ -5848,6 +5894,8 @@ emit_vgpu10_instruction(struct svga_shader_emitter_v10 *emit,
       return emit_issg(emit, inst);
    case TGSI_OPCODE_TEX:
       return emit_tex(emit, inst);
+   case TGSI_OPCODE_TG4:
+      return emit_tg4(emit, inst);
    case TGSI_OPCODE_TEX2:
       return emit_tex2(emit, inst);
    case TGSI_OPCODE_TXP: