[Mesa-dev] [PATCH 1/3] gallium/radeon: avoid redundant work with overlapping in/out arrays

[Nicolai Hähnle]

From: Nicolai Hähnle <nicolai.haehnle at amd.com>

---
 src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
index 8e364c9..da4a030 100644
--- a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
+++ b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
@@ -673,21 +673,22 @@ static void emit_declaration(struct lp_build_tgsi_context *bld_base,
 				}
 				ctx->temps[first * TGSI_NUM_CHANNELS + i] = ptr;
 			}
 		}
 		break;
 	}
 	case TGSI_FILE_INPUT:
 	{
 		unsigned idx;
 		for (idx = decl->Range.First; idx <= decl->Range.Last; idx++) {
-			if (ctx->load_input) {
+			if (ctx->load_input &&
+			    ctx->input_decls[idx].Declaration.File != TGSI_FILE_INPUT) {
 				ctx->input_decls[idx] = *decl;
 
 				if (bld_base->info->processor != PIPE_SHADER_FRAGMENT)
 					ctx->load_input(ctx, idx, decl,
 							&ctx->inputs[idx * 4]);
 			}
 		}
 	}
 	break;
 
@@ -699,20 +700,22 @@ static void emit_declaration(struct lp_build_tgsi_context *bld_base,
 		}
 	}
 	break;
 
 	case TGSI_FILE_OUTPUT:
 	{
 		unsigned idx;
 		for (idx = decl->Range.First; idx <= decl->Range.Last; idx++) {
 			unsigned chan;
 			assert(idx < RADEON_LLVM_MAX_OUTPUTS);
+			if (ctx->soa.outputs[idx][0])
+				continue;
 			for (chan = 0; chan < TGSI_NUM_CHANNELS; chan++) {
 				ctx->soa.outputs[idx][chan] = lp_build_alloca_undef(
 					&ctx->gallivm,
 					ctx->soa.bld_base.base.elem_type, "");
 			}
 		}
 		break;
 	}
 
 	case TGSI_FILE_MEMORY:
-- 
2.7.4