summaryrefslogtreecommitdiffstats
path: root/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
diff options
context:
space:
mode:
authorAlexey Bataev <a.bataev@hotmail.com>2019-01-07 15:45:09 +0000
committerAlexey Bataev <a.bataev@hotmail.com>2019-01-07 15:45:09 +0000
commit25d3de8a0ac4d65f2ac9820ba62a372414780649 (patch)
tree13f97a15bf466de00a4eb5cbd199d2bea69cd5fd /clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
parent27429d87860691defb47b8eb266d1d4832847787 (diff)
downloadbcm5719-llvm-25d3de8a0ac4d65f2ac9820ba62a372414780649.tar.gz
bcm5719-llvm-25d3de8a0ac4d65f2ac9820ba62a372414780649.zip
[OPENMP][NVPTX]Reduce number of barriers in reductions.
After the fix for the syncthreads we don't need to generate extra barriers for the parallel reductions. llvm-svn: 350530
Diffstat (limited to 'clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp')
-rw-r--r--clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp7
1 files changed, 0 insertions, 7 deletions
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
index 21911c96f36..403aefbef9a 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
@@ -3124,7 +3124,6 @@ static void emitReductionListCopy(
/// sync
/// if (I am the first warp)
/// Copy smem[thread_id] to my local D
-/// sync
static llvm::Value *emitInterWarpCopyFunction(CodeGenModule &CGM,
ArrayRef<const Expr *> Privates,
QualType ReductionArrayTy,
@@ -3337,12 +3336,6 @@ static llvm::Value *emitInterWarpCopyFunction(CodeGenModule &CGM,
CGF.EmitBlock(W0MergeBB);
- // While warp 0 copies values from transfer medium, all other warps must
- // wait.
- // kmpc_barrier.
- CGM.getOpenMPRuntime().emitBarrierCall(CGF, Loc, OMPD_unknown,
- /*EmitChecks=*/false,
- /*ForceSimpleCall=*/true);
if (NumIters > 1) {
Cnt = Bld.CreateNSWAdd(Cnt, llvm::ConstantInt::get(CGM.IntTy, /*V=*/1));
CGF.EmitStoreOfScalar(Cnt, CntAddr, /*Volatile=*/false, C.IntTy);
OpenPOWER on IntegriCloud