freedreno/nir: sysval fixes

Signed-off-by: Rob Clark <robclark@freedesktop.org>

diff --git a/src/gallium/drivers/freedreno/a4xx/fd4_emit.c b/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
index 7a244d8a4c5e10f17adccacdcb45142e58ca3d04..353873daa05feac4198d3c2305315f52d0b54b8f 100644 (file)
--- a/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
+++ b/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
@@ -302,7 +302,7 @@ fd4_emit_vertex_bufs(struct fd_ringbuffer *ring, struct fd4_emit *emit)
 
        for (i = 0; i < vp->inputs_count; i++) {
                uint8_t semantic = sem2name(vp->inputs[i].semantic);
-               if (semantic == TGSI_SEMANTIC_VERTEXID)
+               if (semantic == TGSI_SEMANTIC_VERTEXID_NOBASE)
                        vertex_regid = vp->inputs[i].regid;
                else if (semantic == TGSI_SEMANTIC_INSTANCEID)
                        instance_regid = vp->inputs[i].regid;

diff --git a/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c b/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c
index e5f6c2a5e83a6c29061244fcd43387862806044e..05e7049ad55221e8b9c9add681f83888aad347eb 100644 (file)
--- a/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c
@@ -1140,6 +1140,7 @@ static void add_sysval_input(struct ir3_compile *ctx, unsigned name,
        so->inputs[n].interpolate = TGSI_INTERPOLATE_CONSTANT;
        so->total_in++;
 
+       ctx->block->ninputs = MAX2(ctx->block->ninputs, r + 1);
        ctx->block->inputs[r] = instr;
 }
 
@@ -1785,16 +1786,22 @@ emit_instructions(struct ir3_compile *ctx)
        unsigned noutputs = exec_list_length(&ctx->s->outputs) * 4;
 
        /* we need to allocate big enough outputs array so that
-        * we can stuff the kill's at the end:
+        * we can stuff the kill's at the end.  Likewise for vtx
+        * shaders, we need to leave room for sysvals:
         */
-       if (ctx->so->type == SHADER_FRAGMENT)
+       if (ctx->so->type == SHADER_FRAGMENT) {
                noutputs += ARRAY_SIZE(ctx->kill);
+       } else if (ctx->so->type == SHADER_VERTEX) {
+               ninputs += 8;
+       }
 
        ctx->block = ir3_block_create(ctx->ir, 0, ninputs, noutputs);
 
-       if (ctx->so->type == SHADER_FRAGMENT)
+       if (ctx->so->type == SHADER_FRAGMENT) {
                ctx->block->noutputs -= ARRAY_SIZE(ctx->kill);
-
+       } else if (ctx->so->type == SHADER_VERTEX) {
+               ctx->block->ninputs -= 8;
+       }
 
        /* for fragment shader, we have a single input register (usually
         * r0.xy) which is used as the base for bary.f varying fetch instrs: