summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorAlexey Bataev <a.bataev@hotmail.com>2018-10-05 15:27:47 +0000
committerAlexey Bataev <a.bataev@hotmail.com>2018-10-05 15:27:47 +0000
commit6bc2732f71e48c2d7a1d6182b6fc2437f533616d (patch)
tree779e24f9f0767a948509531cf3fc67b39e5dd71e
parentfd006c44eeffe1f018f7fda13eacb67a6606ec9c (diff)
downloadbcm5719-llvm-6bc2732f71e48c2d7a1d6182b6fc2437f533616d.tar.gz
bcm5719-llvm-6bc2732f71e48c2d7a1d6182b6fc2437f533616d.zip
[OPENMP][NVPTX] Fix emission of __kmpc_global_thread_num() for non-SPMD
mode. __kmpc_global_thread_num() should be called before initialization of the runtime. llvm-svn: 343857
-rw-r--r--clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp11
-rw-r--r--clang/test/OpenMP/nvptx_teams_codegen.cpp4
2 files changed, 9 insertions, 6 deletions
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
index f0f0a735f8a..b75891e4f21 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
@@ -1083,12 +1083,15 @@ void CGOpenMPRuntimeNVPTX::emitNonSPMDKernel(const OMPExecutableDirective &D,
CGOpenMPRuntimeNVPTX::WorkerFunctionState &WST)
: EST(EST), WST(WST) {}
void Enter(CodeGenFunction &CGF) override {
- static_cast<CGOpenMPRuntimeNVPTX &>(CGF.CGM.getOpenMPRuntime())
- .emitNonSPMDEntryHeader(CGF, EST, WST);
+ auto &RT = static_cast<CGOpenMPRuntimeNVPTX &>(CGF.CGM.getOpenMPRuntime());
+ RT.emitNonSPMDEntryHeader(CGF, EST, WST);
+ // Skip target region initialization.
+ RT.setLocThreadIdInsertPt(CGF, /*AtCurrentPoint=*/true);
}
void Exit(CodeGenFunction &CGF) override {
- static_cast<CGOpenMPRuntimeNVPTX &>(CGF.CGM.getOpenMPRuntime())
- .emitNonSPMDEntryFooter(CGF, EST);
+ auto &RT = static_cast<CGOpenMPRuntimeNVPTX &>(CGF.CGM.getOpenMPRuntime());
+ RT.clearLocThreadIdInsertPt(CGF);
+ RT.emitNonSPMDEntryFooter(CGF, EST);
}
} Action(EST, WST);
CodeGen.setAction(Action);
diff --git a/clang/test/OpenMP/nvptx_teams_codegen.cpp b/clang/test/OpenMP/nvptx_teams_codegen.cpp
index 2c50f037b22..4e3f2674fc5 100644
--- a/clang/test/OpenMP/nvptx_teams_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_teams_codegen.cpp
@@ -105,7 +105,6 @@ int main (int argc, char **argv) {
// CK2: [[AADDR:%.+]] = alloca i{{[0-9]+}},
// CK2: [[BADDR:%.+]] = alloca i{{[0-9]+}},
// CK2: [[ARGCADDR:%.+]] = alloca i{{[0-9]+}},
-// CK2: {{%.+}} = call i32 @__kmpc_global_thread_num(
// CK2: store i{{[0-9]+}} [[A_IN]], i{{[0-9]+}}* [[AADDR]],
// CK2: store i{{[0-9]+}} [[B_IN]], i{{[0-9]+}}* [[BADDR]],
// CK2: store i{{[0-9]+}} [[ARGC_IN]], i{{[0-9]+}}* [[ARGCADDR]],
@@ -117,6 +116,7 @@ int main (int argc, char **argv) {
// CK2-32: [[ARG:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[ARGCADDR]]
// CK2: [[ARGCADDR:%.+]] = getelementptr inbounds %struct.{{.*}}, %struct.{{.*}}* %{{.*}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0
// CK2: store i{{[0-9]+}} [[ARG]], i{{[0-9]+}}* [[ARGCADDR]],
+// CK2: {{%.+}} = call i32 @__kmpc_global_thread_num(
// CK2: store i{{[0-9]+}}* [[ARGCADDR]], i{{[0-9]+}}** [[ARGCADDR_PTR]],
// CK2: [[ARGCADDR_PTR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARGCADDR_PTR]],
// CK2: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[ARGCADDR_PTR_REF]],
@@ -129,7 +129,6 @@ int main (int argc, char **argv) {
// CK2: [[AADDR:%.+]] = alloca i{{[0-9]+}},
// CK2: [[BADDR:%.+]] = alloca i{{[0-9]+}},
// CK2: [[ARGCADDR:%.+]] = alloca i{{[0-9]+}}**,
-// CK2: {{%.+}} = call i32 @__kmpc_global_thread_num(
// CK2: store i{{[0-9]+}} [[A_IN]], i{{[0-9]+}}* [[AADDR]],
// CK2: store i{{[0-9]+}} [[B_IN]], i{{[0-9]+}}* [[BADDR]],
// CK2: store i{{[0-9]+}}** [[ARGC]], i{{[0-9]+}}*** [[ARGCADDR]],
@@ -137,6 +136,7 @@ int main (int argc, char **argv) {
// CK2: [[ARG:%.+]] = load i{{[0-9]+}}**, i{{[0-9]+}}*** [[ARGCADDR]]
// CK2: [[ARGCADDR:%.+]] = getelementptr inbounds %struct.{{.*}}, %struct.{{.*}}* %{{.*}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0
// CK2: store i{{[0-9]+}}** [[ARG]], i{{[0-9]+}}*** [[ARGCADDR]],
+// CK2: {{%.+}} = call i32 @__kmpc_global_thread_num(
// CK2: store i{{[0-9]+}}*** [[ARGCADDR]], i{{[0-9]+}}**** [[ARGCADDR_PTR]],
// CK2: [[ARGCADDR_PTR_REF:%.+]] = load i{{[0-9]+}}***, i{{[0-9]+}}**** [[ARGCADDR_PTR]],
// CK2: store i{{[0-9]+}}** null, i{{[0-9]+}}*** [[ARGCADDR_PTR_REF]],
OpenPOWER on IntegriCloud