R600/SI: Insert s_waitcnt before s_barrier instructions.

This ensures that all memory operations are complete when all threads
reach the barrier.

llvm-svn: 225290

diff --git a/llvm/lib/Target/R600/SIInsertWaits.cpp b/llvm/lib/Target/R600/SIInsertWaits.cpp
index 2e56508..181b116 100644 (file)
--- a/llvm/lib/Target/R600/SIInsertWaits.cpp
+++ b/llvm/lib/Target/R600/SIInsertWaits.cpp
@@ -428,7 +428,11 @@ bool SIInsertWaits::runOnMachineFunction(MachineFunction &MF) {
     for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end();
          I != E; ++I) {
 
-      Changes |= insertWait(MBB, I, handleOperands(*I));
+      // Wait for everything before a barrier.
+      if (I->getOpcode() == AMDGPU::S_BARRIER)
+        Changes |= insertWait(MBB, I, LastIssued);
+      else
+        Changes |= insertWait(MBB, I, handleOperands(*I));
       pushInstruction(MBB, I);
     }
 

diff --git a/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.global.ll b/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.global.ll
index 31e1709..a11d9ae 100644 (file)
--- a/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.global.ll
+++ b/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.global.ll
@@ -3,6 +3,8 @@
 
 ; FUNC-LABEL: {{^}}test_barrier_global:
 ; EG: GROUP_BARRIER
+; SI: buffer_store_dword
+; SI: s_waitcnt
 ; SI: s_barrier
 
 define void @test_barrier_global(i32 addrspace(1)* %out) {

diff --git a/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.local.ll b/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.local.ll
index 94b3cbc..76c2453 100644 (file)
--- a/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.local.ll
+++ b/llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.local.ll
@@ -3,6 +3,9 @@
 
 ; FUNC-LABEL: {{^}}test_barrier_local:
 ; EG: GROUP_BARRIER
+
+; SI: buffer_store_dword
+; SI: s_waitcnt
 ; SI: s_barrier
 
 define void @test_barrier_local(i32 addrspace(1)* %out) {