maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

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
This commit is contained in:
Tom Stellard 2015-01-06 19:52:07 +00:00
parent b3931b814a
commit 9d6797ae58
3 changed files with 10 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/R600/SIInsertWaits.cpp
View File

@ -428,6 +428,10 @@ bool SIInsertWaits::runOnMachineFunction(MachineFunction &MF) {
for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end(); for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end();
I != E; ++I) { I != E; ++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)); Changes |= insertWait(MBB, I, handleOperands(*I));
pushInstruction(MBB, I); pushInstruction(MBB, I);
} }

2
llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.global.ll
View File

@ -3,6 +3,8 @@
; FUNC-LABEL: {{^}}test_barrier_global: ; FUNC-LABEL: {{^}}test_barrier_global:
; EG: GROUP_BARRIER ; EG: GROUP_BARRIER
; SI: buffer_store_dword
; SI: s_waitcnt
; SI: s_barrier ; SI: s_barrier
define void @test_barrier_global(i32 addrspace(1)* %out) { define void @test_barrier_global(i32 addrspace(1)* %out) {

3
llvm/test/CodeGen/R600/llvm.AMDGPU.barrier.local.ll
View File

@ -3,6 +3,9 @@
; FUNC-LABEL: {{^}}test_barrier_local: ; FUNC-LABEL: {{^}}test_barrier_local:
; EG: GROUP_BARRIER ; EG: GROUP_BARRIER
; SI: buffer_store_dword
; SI: s_waitcnt
; SI: s_barrier ; SI: s_barrier
define void @test_barrier_local(i32 addrspace(1)* %out) { define void @test_barrier_local(i32 addrspace(1)* %out) {