Skip to content

libcu++: respect atomic scopes in pointer arithmetic operations. #10809

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
griwes merged 1 commit into from
Aug 14, 2026
+62 −2
Merged

libcu++: respect atomic scopes in pointer arithmetic operations. #10809

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions libcudacxx/include/cuda/std/__atomic/api/common.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -163,12 +163,12 @@
_CCCL_HOST_DEVICE_API inline _Tp fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst) \
_CONST _VOLATILE noexcept \
{ \
return __atomic_fetch_add_dispatch(&__a, __op, __m, __thread_scope_system_tag{}); \
return __atomic_fetch_add_dispatch(&__a, __op, __m, _Sco{}); \
} \
_CCCL_HOST_DEVICE_API inline _Tp fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst) \
_CONST _VOLATILE noexcept \
{ \
return __atomic_fetch_sub_dispatch(&__a, __op, __m, __thread_scope_system_tag{}); \
return __atomic_fetch_sub_dispatch(&__a, __op, __m, _Sco{}); \
} \
_CCCL_HOST_DEVICE_API inline _Tp operator++(int) _CONST _VOLATILE noexcept \
{ \
Expand Down
30 changes: 30 additions & 0 deletions libcudacxx/test/atomic_codegen/atomic_add_non_volatile.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,24 @@ __global__ void add_relaxed_device_non_volatile(int* data, int* out, int n)
*out = ref.fetch_add(n, cuda::std::memory_order_relaxed);
}

__global__ void add_relaxed_block_pointer_non_volatile(int** data, int** out, int n)
{
auto ref = cuda::atomic_ref<int*, cuda::thread_scope_block>{*data};
*out = ref.fetch_add(n, cuda::std::memory_order_relaxed);
Comment thread
coderabbitai[bot] marked this conversation as resolved.
}

__global__ void add_relaxed_device_pointer_non_volatile(int** data, int** out, int n)
{
auto ref = cuda::atomic_ref<int*, cuda::thread_scope_device>{*data};
*out = ref.fetch_add(n, cuda::std::memory_order_relaxed);
}

__global__ void add_relaxed_system_pointer_non_volatile(int** data, int** out, int n)
{
auto ref = cuda::atomic_ref<int*, cuda::thread_scope_system>{*data};
*out = ref.fetch_add(n, cuda::std::memory_order_relaxed);
}

/*

; SM8X-LABEL: .target sm_80
Expand All @@ -18,4 +36,16 @@ __global__ void add_relaxed_device_non_volatile(int* data, int* out, int n)
; SM8X-NEXT: st.global.{{b|u}}32 [%rd[[#GOUT]]], %r[[#DEST]];
; SM8X-NEXT: ret;

; SM8X-LABEL: .visible .entry {{_.*add_relaxed_block_pointer_non_volatile.*}}(
; SM8X: {{.*}}atom.add.relaxed.cta.u64{{.*}}
; SM8X: ret;

; SM8X-LABEL: .visible .entry {{_.*add_relaxed_device_pointer_non_volatile.*}}(
; SM8X: {{.*}}atom.add.relaxed.gpu.u64{{.*}}
; SM8X: ret;

; SM8X-LABEL: .visible .entry {{_.*add_relaxed_system_pointer_non_volatile.*}}(
; SM8X: {{.*}}atom.add.relaxed.sys.u64{{.*}}
; SM8X: ret;

*/
30 changes: 30 additions & 0 deletions libcudacxx/test/atomic_codegen/atomic_sub_non_volatile.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,24 @@ __global__ void sub_relaxed_device_non_volatile(int* data, int* out, int n)
*out = ref.fetch_sub(n, cuda::std::memory_order_relaxed);
}

__global__ void sub_relaxed_block_pointer_non_volatile(int** data, int** out, int n)
{
auto ref = cuda::atomic_ref<int*, cuda::thread_scope_block>{*data};
*out = ref.fetch_sub(n, cuda::std::memory_order_relaxed);
}

__global__ void sub_relaxed_device_pointer_non_volatile(int** data, int** out, int n)
{
auto ref = cuda::atomic_ref<int*, cuda::thread_scope_device>{*data};
*out = ref.fetch_sub(n, cuda::std::memory_order_relaxed);
}

__global__ void sub_relaxed_system_pointer_non_volatile(int** data, int** out, int n)
{
auto ref = cuda::atomic_ref<int*, cuda::thread_scope_system>{*data};
*out = ref.fetch_sub(n, cuda::std::memory_order_relaxed);
}

/*

; SM8X-LABEL: .target sm_80
Expand All @@ -19,4 +37,16 @@ __global__ void sub_relaxed_device_non_volatile(int* data, int* out, int n)
; SM8X-NEXT: st.global.{{b|u}}32 [%rd[[#GOUT]]], %r[[#DEST]];
; SM8X-NEXT: ret;

; SM8X-LABEL: .visible .entry {{_.*sub_relaxed_block_pointer_non_volatile.*}}(
; SM8X: {{.*}}atom.add.relaxed.cta.u64{{.*}}
; SM8X: ret;

; SM8X-LABEL: .visible .entry {{_.*sub_relaxed_device_pointer_non_volatile.*}}(
; SM8X: {{.*}}atom.add.relaxed.gpu.u64{{.*}}
; SM8X: ret;

; SM8X-LABEL: .visible .entry {{_.*sub_relaxed_system_pointer_non_volatile.*}}(
; SM8X: {{.*}}atom.add.relaxed.sys.u64{{.*}}
; SM8X: ret;

*/
Loading