r600g: gpu_shader5 gl_SampleMaskIn support

Map TGSI_SEMANTIC_SAMPLEMASK to register/component.
Enable face register when sample mask is needed by shader.
Requires Evergreen/Cayman

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/docs/GL3.txt b/docs/GL3.txt
index 89a4ea31613db61fa9168eee464f04268a96c763..59b20dc2e425684dbf555a7a7dea0ec4acc4dbde 100644 (file)
--- a/docs/GL3.txt
+++ b/docs/GL3.txt
@@ -109,7 +109,7 @@ GL 4.0:
   - Enhanced textureGather                             DONE (i965, nvc0, radeonsi)
   - Geometry shader instancing                         DONE (i965, nvc0)
   - Geometry shader multiple streams                   DONE (i965, nvc0)
-  - Enhanced per-sample shading                        DONE (i965)
+  - Enhanced per-sample shading                        DONE (i965, r600)
   - Interpolation functions                            DONE (i965)
   - New overload resolution rules                      DONE
   GL_ARB_gpu_shader_fp64                               started (Dave)

diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 45c611e5cc24ba11197d3a17be5926e30e5d9a64..608b05b332fe76238a16386e08878c225c705637 100644 (file)
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -2841,8 +2841,14 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
                   POSITION goes via GPRs from the SC so isn't counted */
                if (rshader->input[i].name == TGSI_SEMANTIC_POSITION)
                        pos_index = i;
-               else if (rshader->input[i].name == TGSI_SEMANTIC_FACE)
-                       face_index = i;
+               else if (rshader->input[i].name == TGSI_SEMANTIC_FACE) {
+                       if (face_index == -1)
+                               face_index = i;
+               }
+               else if (rshader->input[i].name == TGSI_SEMANTIC_SAMPLEMASK) {
+                       if (face_index == -1)
+                               face_index = i; /* lives in same register, same enable bit */
+               }
                else {
                        ninterp++;
                        if (rshader->input[i].interpolate == TGSI_INTERPOLATE_LINEAR)

diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 175c09b9501191b903d41b92e91b56bd6f7157db..00b2f4aeaed1cf7c26f6e7d5808fb99221ebe686 100644 (file)
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -287,7 +287,9 @@ struct r600_shader_ctx {
        boolean                                 input_linear;
        boolean                                 input_perspective;
        int                                     num_interp_gpr;
+       /* evergreen/cayman also store sample mask in face register */
        int                                     face_gpr;
+       boolean                                 has_samplemask;
        int                                     colors_used;
        boolean                 clip_vertex_write;
        unsigned                cv_output;
@@ -498,7 +500,8 @@ static int r600_spi_sid(struct r600_shader_io * io)
        if (name == TGSI_SEMANTIC_POSITION ||
            name == TGSI_SEMANTIC_PSIZE ||
            name == TGSI_SEMANTIC_EDGEFLAG ||
-           name == TGSI_SEMANTIC_FACE)
+           name == TGSI_SEMANTIC_FACE ||
+           name == TGSI_SEMANTIC_SAMPLEMASK)
                index = 0;
        else {
                if (name == TGSI_SEMANTIC_GENERIC) {
@@ -585,7 +588,8 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx)
                        ctx->shader->input[i].spi_sid = r600_spi_sid(&ctx->shader->input[i]);
                        switch (ctx->shader->input[i].name) {
                        case TGSI_SEMANTIC_FACE:
-                               ctx->face_gpr = ctx->shader->input[i].gpr;
+                               if (ctx->face_gpr == -1)
+                                       ctx->face_gpr = ctx->shader->input[i].gpr;
                                break;
                        case TGSI_SEMANTIC_COLOR:
                                ctx->colors_used++;
@@ -675,7 +679,14 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx)
                break;
 
        case TGSI_FILE_SYSTEM_VALUE:
-               if (d->Semantic.Name == TGSI_SEMANTIC_INSTANCEID) {
+               if (d->Semantic.Name == TGSI_SEMANTIC_SAMPLEMASK) {
+                       ctx->has_samplemask = true;
+                       /* lives in Front Face GPR */
+                       if (ctx->face_gpr == -1)
+                               ctx->face_gpr = ctx->file_offset[TGSI_FILE_SYSTEM_VALUE] + d->Range.First;
+                       break;
+               }
+               else if (d->Semantic.Name == TGSI_SEMANTIC_INSTANCEID) {
                        if (!ctx->native_integers) {
                                struct r600_bytecode_alu alu;
                                memset(&alu, 0, sizeof(struct r600_bytecode_alu));
@@ -729,7 +740,8 @@ static int evergreen_gpr_count(struct r600_shader_ctx *ctx)
        for (i = 0; i < ctx->info.num_inputs; i++) {
                /* skip position/face */
                if (ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_POSITION ||
-                   ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_FACE)
+                   ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_FACE ||
+                   ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_SAMPLEMASK)
                        continue;
                if (ctx->info.input_interpolate[i] == TGSI_INTERPOLATE_LINEAR)
                        ctx->input_linear = TRUE;
@@ -781,7 +793,13 @@ static void tgsi_src(struct r600_shader_ctx *ctx,
                r600_src->sel = V_SQ_ALU_SRC_LITERAL;
                memcpy(r600_src->value, ctx->literals + index * 4, sizeof(r600_src->value));
        } else if (tgsi_src->Register.File == TGSI_FILE_SYSTEM_VALUE) {
-               if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_INSTANCEID) {
+               if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_SAMPLEMASK) {
+                       r600_src->swizzle[0] = 2; // Z value
+                       r600_src->swizzle[0] = 2;
+                       r600_src->swizzle[0] = 2;
+                       r600_src->swizzle[0] = 2;
+                       r600_src->sel = ctx->face_gpr;
+               } else if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_INSTANCEID) {
                        r600_src->swizzle[0] = 3;
                        r600_src->swizzle[1] = 3;
                        r600_src->swizzle[2] = 3;
@@ -1585,6 +1603,7 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
        ctx.gs_next_vertex = 0;
 
        ctx.face_gpr = -1;
+       ctx.has_samplemask = false;
        ctx.fragcoord_input = -1;
        ctx.colors_used = 0;
        ctx.clip_vertex_write = 0;
@@ -1744,6 +1763,14 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
        
        shader->ring_item_size = ctx.next_ring_offset;
 
+       /* Need to tell setup to program FACE register */
+       if (ctx.has_samplemask && ctx.face_gpr != -1) {
+               i = ctx.shader->ninput++;
+               ctx.shader->input[i].name = TGSI_SEMANTIC_SAMPLEMASK;
+               ctx.shader->input[i].spi_sid = 0;
+               ctx.shader->input[i].gpr = ctx.face_gpr;
+       }
+
        /* Process two side if needed */
        if (shader->two_side && ctx.colors_used) {
                int i, count = ctx.shader->ninput;