[Mesa-dev] [PATCH 10/10] radeonsi: wait for outstanding LDS instructions in memory barriers if needed

[Marek Olšák]

From: Marek Olšák <marek.olsak at amd.com>

Cc: 13.0 12.0 <mesa-stable at lists.freedesktop.org>
---
 src/gallium/drivers/radeonsi/si_shader.c | 18 +++++++++++++++++-
 1 file changed, 17 insertions(+), 1 deletion(-)

diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 45896bd..dc5c67a 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -3187,42 +3187,58 @@ static void build_tex_intrinsic(const struct lp_build_tgsi_action *action,
 static void emit_optimization_barrier(struct si_shader_context *ctx)
 {
 	LLVMBuilderRef builder = ctx->gallivm.builder;
 	LLVMTypeRef ftype = LLVMFunctionType(ctx->voidt, NULL, 0, false);
 	LLVMValueRef inlineasm = LLVMConstInlineAsm(ftype, "", "", true, false);
 	LLVMBuildCall(builder, inlineasm, NULL, 0, "");
 }
 #endif
 
 /* Combine these with & instead of |. */
+#define NOOP_WAITCNT 0xf7f
 #define LGKM_CNT 0x07f
 #define VM_CNT 0xf70
 
 static void emit_waitcnt(struct si_shader_context *ctx, unsigned simm16)
 {
 	struct gallivm_state *gallivm = &ctx->gallivm;
 	LLVMBuilderRef builder = gallivm->builder;
 	LLVMValueRef args[1] = {
 		lp_build_const_int32(gallivm, simm16)
 	};
 	lp_build_intrinsic(builder, "llvm.amdgcn.s.waitcnt",
 			   ctx->voidt, args, 1, 0);
 }
 
 static void membar_emit(
 		const struct lp_build_tgsi_action *action,
 		struct lp_build_tgsi_context *bld_base,
 		struct lp_build_emit_data *emit_data)
 {
 	struct si_shader_context *ctx = si_shader_context(bld_base);
+	LLVMValueRef src0 = lp_build_emit_fetch(bld_base, emit_data->inst, 0, 0);
+	unsigned flags = LLVMConstIntGetZExtValue(src0);
+	unsigned waitcnt = NOOP_WAITCNT;
 
-	emit_waitcnt(ctx, VM_CNT);
+	if (flags & TGSI_MEMBAR_THREAD_GROUP)
+		waitcnt &= VM_CNT & LGKM_CNT;
+
+	if (flags & (TGSI_MEMBAR_ATOMIC_BUFFER |
+		     TGSI_MEMBAR_SHADER_BUFFER |
+		     TGSI_MEMBAR_SHADER_IMAGE))
+		waitcnt &= VM_CNT;
+
+	if (flags & TGSI_MEMBAR_SHARED)
+		waitcnt &= LGKM_CNT;
+
+	if (waitcnt != NOOP_WAITCNT)
+		emit_waitcnt(ctx, waitcnt);
 }
 
 static LLVMValueRef
 shader_buffer_fetch_rsrc(struct si_shader_context *ctx,
 			 const struct tgsi_full_src_register *reg)
 {
 	LLVMValueRef index;
 	LLVMValueRef rsrc_ptr = LLVMGetParam(ctx->main_fn,
 					     SI_PARAM_SHADER_BUFFERS);
 
-- 
2.7.4