[clang] [Clang][NVPTX] fix `no member named 'this_cluster'` (PR #187039)
via cfe-commits
cfe-commits at lists.llvm.org
Thu Apr 30 13:58:13 PDT 2026
https://github.com/NeKon69 updated https://github.com/llvm/llvm-project/pull/187039
>From f4d922cd241cc428d12ff6abf7948f8c29650d2b Mon Sep 17 00:00:00 2001
From: NeKon69 <nobodqwe at gmail.com>
Date: Tue, 17 Mar 2026 17:46:21 +0300
Subject: [PATCH 1/5] [NVPTX] add a macro to correctly resolve
`cg::this_cluster`
---
clang/lib/Basic/Targets/NVPTX.cpp | 6 ++++++
clang/test/Preprocessor/predefined-arch-macros.c | 6 ++++++
2 files changed, 12 insertions(+)
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index 6526eeff3d718..5b4286f8bfd5e 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -184,6 +184,12 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
unsigned ArchID = CudaArchToID(GPU);
Builder.defineMacro("__CUDA_ARCH__", llvm::Twine(ArchID));
+ // Per
+ // https://docs.nvidia.com/cuda/cuda-programming-guide/04-special-topics/cooperative-groups.html#id4
+ // clusters are only available with >=9.0 compute capability
+ if (ArchID >= 900)
+ Builder.defineMacro("_CG_CLUSTER_INTRINSICS_AVAILABLE");
+
if (IsNVIDIAAcceleratedOffloadArch(GPU))
Builder.defineMacro(
"__CUDA_ARCH_FEAT_SM" + llvm::Twine(ArchID / 10) + "_ALL", "1");
diff --git a/clang/test/Preprocessor/predefined-arch-macros.c b/clang/test/Preprocessor/predefined-arch-macros.c
index cb2d13d59d8bf..7c2139c2f6a40 100644
--- a/clang/test/Preprocessor/predefined-arch-macros.c
+++ b/clang/test/Preprocessor/predefined-arch-macros.c
@@ -4602,11 +4602,17 @@
// RUN: %clang -march=sm_75 -E -dM %s -o - 2>&1 \
// RUN: -target nvptx64-unknown-unknown \
// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_SM_75
+// RUN: %clang -march=sm_90 -E -dM %s -o - 2>&1 \
+// RUN: -target nvptx64-unknown-unknown \
+// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_SM_90
// RUN: %clang -E -dM %s -o - 2>&1 \
// RUN: -target nvptx64-unknown-unknown \
// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_UNSET
// CHECK_ARCH_SM_75: #define __CUDA_ARCH__ 750
+// CHECK_ARCH_SM_90: #define __CUDA_ARCH__ 900
+// CHECK_ARCH_SM_90: #define _CG_CLUSTER_INTRINSICS_AVAILABLE 1
// CHECK_ARCH_UNSET-NOT: #define __CUDA_ARCH__
+// CHECK_ARCH_UNSET-NOT: #define _CG_CLUSTER_INTRINSICS_AVAILABLE
// CHECK_NVPTX: #define __NVPTX__ 1
// Begin amdgcn tests ----------------
>From 0ed5821be47e21d9e8c65326364bffc049f62e23 Mon Sep 17 00:00:00 2001
From: NeKon69 <nobodqwe at gmail.com>
Date: Tue, 17 Mar 2026 19:39:44 +0300
Subject: [PATCH 2/5] reorder checks
---
clang/test/Preprocessor/predefined-arch-macros.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/test/Preprocessor/predefined-arch-macros.c b/clang/test/Preprocessor/predefined-arch-macros.c
index 7c2139c2f6a40..fcb9cad0b7e63 100644
--- a/clang/test/Preprocessor/predefined-arch-macros.c
+++ b/clang/test/Preprocessor/predefined-arch-macros.c
@@ -4609,8 +4609,8 @@
// RUN: -target nvptx64-unknown-unknown \
// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_UNSET
// CHECK_ARCH_SM_75: #define __CUDA_ARCH__ 750
-// CHECK_ARCH_SM_90: #define __CUDA_ARCH__ 900
// CHECK_ARCH_SM_90: #define _CG_CLUSTER_INTRINSICS_AVAILABLE 1
+// CHECK_ARCH_SM_90: #define __CUDA_ARCH__ 900
// CHECK_ARCH_UNSET-NOT: #define __CUDA_ARCH__
// CHECK_ARCH_UNSET-NOT: #define _CG_CLUSTER_INTRINSICS_AVAILABLE
// CHECK_NVPTX: #define __NVPTX__ 1
>From 9c56c8d30afdab60f713d3fd28dedf0ebc7a26eb Mon Sep 17 00:00:00 2001
From: NeKon69 <nobodqwe at gmail.com>
Date: Thu, 30 Apr 2026 23:22:50 +0300
Subject: [PATCH 3/5] move macro definition
---
clang/lib/Basic/Targets/NVPTX.cpp | 6 ------
clang/lib/Headers/__clang_cuda_runtime_wrapper.h | 2 ++
2 files changed, 2 insertions(+), 6 deletions(-)
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index 5b4286f8bfd5e..6526eeff3d718 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -184,12 +184,6 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
unsigned ArchID = CudaArchToID(GPU);
Builder.defineMacro("__CUDA_ARCH__", llvm::Twine(ArchID));
- // Per
- // https://docs.nvidia.com/cuda/cuda-programming-guide/04-special-topics/cooperative-groups.html#id4
- // clusters are only available with >=9.0 compute capability
- if (ArchID >= 900)
- Builder.defineMacro("_CG_CLUSTER_INTRINSICS_AVAILABLE");
-
if (IsNVIDIAAcceleratedOffloadArch(GPU))
Builder.defineMacro(
"__CUDA_ARCH_FEAT_SM" + llvm::Twine(ArchID / 10) + "_ALL", "1");
diff --git a/clang/lib/Headers/__clang_cuda_runtime_wrapper.h b/clang/lib/Headers/__clang_cuda_runtime_wrapper.h
index 295f4191f9927..cf0fd3efcca6a 100644
--- a/clang/lib/Headers/__clang_cuda_runtime_wrapper.h
+++ b/clang/lib/Headers/__clang_cuda_runtime_wrapper.h
@@ -518,5 +518,7 @@ extern "C" unsigned __cudaPushCallConfiguration(dim3 gridDim, dim3 blockDim,
void *stream = 0);
#endif
+#define _CG_CLUSTER_INTRINSICS_AVAILABLE 1
+
#endif // __CUDA__
#endif // __CLANG_CUDA_RUNTIME_WRAPPER_H__
>From aa12ceb22f08a056dd4f4a57056d2681a596d2f7 Mon Sep 17 00:00:00 2001
From: NeKon69 <nobodqwe at gmail.com>
Date: Thu, 30 Apr 2026 23:53:50 +0300
Subject: [PATCH 4/5] move tests
---
clang/test/Headers/cuda_wrapper_cluster_intrinsics.cpp | 10 ++++++++++
clang/test/Preprocessor/predefined-arch-macros.c | 2 --
2 files changed, 10 insertions(+), 2 deletions(-)
create mode 100644 clang/test/Headers/cuda_wrapper_cluster_intrinsics.cpp
diff --git a/clang/test/Headers/cuda_wrapper_cluster_intrinsics.cpp b/clang/test/Headers/cuda_wrapper_cluster_intrinsics.cpp
new file mode 100644
index 0000000000000..e58a71d623208
--- /dev/null
+++ b/clang/test/Headers/cuda_wrapper_cluster_intrinsics.cpp
@@ -0,0 +1,10 @@
+// RUN: %clang_cc1 -E -dM -x cuda %s \
+// RUN: | FileCheck -match-full-lines %s --check-prefix=NO-WRAPPER
+// RUN: %clang_cc1 -E -dM -x cuda \
+// RUN: -include __clang_cuda_runtime_wrapper.h \
+// RUN: -internal-isystem %S/../../lib/Headers \
+// RUN: -internal-isystem %S/Inputs/include \
+// RUN: %s | FileCheck -match-full-lines %s --check-prefix=WRAPPER
+
+// NO-WRAPPER-NOT: #define _CG_CLUSTER_INTRINSICS_AVAILABLE 1
+// WRAPPER: #define _CG_CLUSTER_INTRINSICS_AVAILABLE 1
diff --git a/clang/test/Preprocessor/predefined-arch-macros.c b/clang/test/Preprocessor/predefined-arch-macros.c
index fcb9cad0b7e63..3c4d82c90f77f 100644
--- a/clang/test/Preprocessor/predefined-arch-macros.c
+++ b/clang/test/Preprocessor/predefined-arch-macros.c
@@ -4609,10 +4609,8 @@
// RUN: -target nvptx64-unknown-unknown \
// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_UNSET
// CHECK_ARCH_SM_75: #define __CUDA_ARCH__ 750
-// CHECK_ARCH_SM_90: #define _CG_CLUSTER_INTRINSICS_AVAILABLE 1
// CHECK_ARCH_SM_90: #define __CUDA_ARCH__ 900
// CHECK_ARCH_UNSET-NOT: #define __CUDA_ARCH__
-// CHECK_ARCH_UNSET-NOT: #define _CG_CLUSTER_INTRINSICS_AVAILABLE
// CHECK_NVPTX: #define __NVPTX__ 1
// Begin amdgcn tests ----------------
>From 880b0eb1e42dd1a24a0cc2c40b7d20b592f72ccf Mon Sep 17 00:00:00 2001
From: NeKon69 <nobodqwe at gmail.com>
Date: Thu, 30 Apr 2026 23:57:57 +0300
Subject: [PATCH 5/5] remove stale checks
---
clang/test/Preprocessor/predefined-arch-macros.c | 4 ----
1 file changed, 4 deletions(-)
diff --git a/clang/test/Preprocessor/predefined-arch-macros.c b/clang/test/Preprocessor/predefined-arch-macros.c
index 3c4d82c90f77f..cb2d13d59d8bf 100644
--- a/clang/test/Preprocessor/predefined-arch-macros.c
+++ b/clang/test/Preprocessor/predefined-arch-macros.c
@@ -4602,14 +4602,10 @@
// RUN: %clang -march=sm_75 -E -dM %s -o - 2>&1 \
// RUN: -target nvptx64-unknown-unknown \
// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_SM_75
-// RUN: %clang -march=sm_90 -E -dM %s -o - 2>&1 \
-// RUN: -target nvptx64-unknown-unknown \
-// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_SM_90
// RUN: %clang -E -dM %s -o - 2>&1 \
// RUN: -target nvptx64-unknown-unknown \
// RUN: | FileCheck -match-full-lines %s -check-prefixes=CHECK_NVPTX,CHECK_ARCH_UNSET
// CHECK_ARCH_SM_75: #define __CUDA_ARCH__ 750
-// CHECK_ARCH_SM_90: #define __CUDA_ARCH__ 900
// CHECK_ARCH_UNSET-NOT: #define __CUDA_ARCH__
// CHECK_NVPTX: #define __NVPTX__ 1
More information about the cfe-commits
mailing list