replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[OpenMP] Add function attribute for triggering data sharing. 

Summary:
The backend should only emit data sharing code for the cases where it is needed.
A new function attribute is used by Clang to enable data sharing only for the cases where OpenMP semantics require it and there are variables that need to be shared.

Reviewers: hfinkel, Hahnfeld, ABataev, carlo.bertolli, caomhin

Reviewed By: ABataev

Subscribers: cfe-commits, jholewinski

Differential Revision: https://reviews.llvm.org/D41123

llvm-svn: 320527
This commit is contained in:
Gheorghe-Teodor Bercea 2017-12-12 21:38:43 +00:00
parent 338c4b2336
commit b4c74c6603
2 changed files with 10 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
View File

@ -942,6 +942,8 @@ void CGOpenMPRuntimeNVPTX::emitGenericParallelCall(
llvm::Value *ID = Bld.CreateBitOrPointerCast(WFn, CGM.Int8PtrTy); llvm::Value *ID = Bld.CreateBitOrPointerCast(WFn, CGM.Int8PtrTy);
if (!CapturedVars.empty()) { if (!CapturedVars.empty()) {
// There's somehting to share, add the attribute
CGF.CurFn->addFnAttr("has-nvptx-shared-depot");
// Prepare for parallel region. Indicate the outlined function. // Prepare for parallel region. Indicate the outlined function.
Address SharedArgs = Address SharedArgs =
CGF.CreateDefaultAlignTempAlloca(CGF.VoidPtrPtrTy, CGF.CreateDefaultAlignTempAlloca(CGF.VoidPtrPtrTy,

11
clang/test/OpenMP/nvptx_data_sharing.cpp
View File

@ -22,7 +22,7 @@ void test_ds(){
/// ========= In the worker function ========= /// /// ========= In the worker function ========= ///
// CK1: define internal void @__omp_offloading_{{.*}}test_ds{{.*}}worker(){{.*}}{ // CK1: define internal void @__omp_offloading_{{.*}}test_ds{{.*}}worker() [[ATTR1:#.*]] {
// CK1: [[SHAREDARGS:%.+]] = alloca i8** // CK1: [[SHAREDARGS:%.+]] = alloca i8**
// CK1: call i1 @__kmpc_kernel_parallel(i8** %work_fn, i8*** [[SHAREDARGS]]) // CK1: call i1 @__kmpc_kernel_parallel(i8** %work_fn, i8*** [[SHAREDARGS]])
// CK1: [[SHARGSTMP:%.+]] = load i8**, i8*** [[SHAREDARGS]] // CK1: [[SHARGSTMP:%.+]] = load i8**, i8*** [[SHAREDARGS]]
@ -30,7 +30,7 @@ void test_ds(){
/// ========= In the kernel function ========= /// /// ========= In the kernel function ========= ///
// CK1: {{.*}}define void @__omp_offloading{{.*}}test_ds{{.*}}() // CK1: {{.*}}define void @__omp_offloading{{.*}}test_ds{{.*}}() [[ATTR2:#.*]] {
// CK1: [[SHAREDARGS1:%.+]] = alloca i8** // CK1: [[SHAREDARGS1:%.+]] = alloca i8**
// CK1: call void @__kmpc_kernel_prepare_parallel({{.*}}, i8*** [[SHAREDARGS1]], i32 1) // CK1: call void @__kmpc_kernel_prepare_parallel({{.*}}, i8*** [[SHAREDARGS1]], i32 1)
// CK1: [[SHARGSTMP1:%.+]] = load i8**, i8*** [[SHAREDARGS1]] // CK1: [[SHARGSTMP1:%.+]] = load i8**, i8*** [[SHAREDARGS1]]
@ -40,7 +40,7 @@ void test_ds(){
/// ========= In the data sharing wrapper function ========= /// /// ========= In the data sharing wrapper function ========= ///
// CK1: {{.*}}define internal void @__omp_outlined___wrapper({{.*}}i8**){{.*}}{ // CK1: {{.*}}define internal void @__omp_outlined___wrapper({{.*}}i8**) [[ATTR1]] {
// CK1: [[SHAREDARGS2:%.+]] = alloca i8** // CK1: [[SHAREDARGS2:%.+]] = alloca i8**
// CK1: store i8** %2, i8*** [[SHAREDARGS2]] // CK1: store i8** %2, i8*** [[SHAREDARGS2]]
// CK1: [[SHARGSTMP3:%.+]] = load i8**, i8*** [[SHAREDARGS2]] // CK1: [[SHARGSTMP3:%.+]] = load i8**, i8*** [[SHAREDARGS2]]
@ -49,4 +49,9 @@ void test_ds(){
// CK1: [[SHARGSTMP6:%.+]] = load i32*, i32** [[SHARGSTMP5]] // CK1: [[SHARGSTMP6:%.+]] = load i32*, i32** [[SHARGSTMP5]]
// CK1: call void @__omp_outlined__({{.*}}, i32* [[SHARGSTMP6]]) // CK1: call void @__omp_outlined__({{.*}}, i32* [[SHARGSTMP6]])
/// ========= Attributes ========= ///
// CK1-NOT: attributes [[ATTR1]] = { {{.*}}"has-nvptx-shared-depot"{{.*}} }
// CK1: attributes [[ATTR2]] = { {{.*}}"has-nvptx-shared-depot"{{.*}} }
#endif #endif