Skip to content

[sycl_ext_oneapi_clock] implement NVPTX case #21280

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

Open
tdavidcl wants to merge 4 commits into
base: sycl
Choose a base branch
from
+12 −3
Open

[sycl_ext_oneapi_clock] implement NVPTX case #21280

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f6ed79d
[sycl_ext_oneapi_clock] implement NVPTX case
tdavidcl Feb 12, 2026
064f8ef
typo
tdavidcl Feb 12, 2026
ba65f4a
Update sycl/include/sycl/ext/oneapi/experimental/clock.hpp
tdavidcl Feb 12, 2026
b43c2b8
use __builtin_readcyclecounter as fallback
tdavidcl Feb 13, 2026
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
15 changes: 12 additions & 3 deletions sycl/include/sycl/ext/oneapi/experimental/clock.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,9 +26,18 @@ enum class clock_scope : int {
namespace detail {
template <clock_scope Scope> inline uint64_t clock_impl() {
#ifdef __SYCL_DEVICE_ONLY__
#if defined(__NVPTX__) || defined(__AMDGCN__)
// Currently clock() is not supported on NVPTX and AMDGCN.
return 0;
// here note that __builtin_readcyclecounter is used as fallback.
// this is due to potential higher overhead compared to a native API call
// see : https://github.com/ROCm/ROCm/issues/1288
#if defined(__NVPTX__)
if constexpr (Scope == work_group || Scope == sub_group) {
Copy link
Contributor

@KornevNikita KornevNikita Feb 16, 2026 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
if constexpr (Scope == work_group || Scope == sub_group) {
if constexpr (Scope == clock_scope::work_group || Scope == clock_scope::sub_group) {

Note - do not apply this as is, clang-format will fail because strings should be <= 80 symbols.

Copy link
Contributor

@KornevNikita KornevNikita Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

probably like:

if constexpr (Scope == clock_scope::work_group ||
              Scope == clock_scope::sub_group) {

return __nvvm_read_ptx_sreg_clock64();
} else {
return __builtin_readcyclecounter();
}
#elif defined(__AMDGCN__)
// No direct variant of clock() is currently implemented for AMDGCN
return __builtin_readcyclecounter();
#else
return __spirv_ReadClockKHR(static_cast<int>(Scope));
#endif // defined(__NVPTX__) || defined(__AMDGCN__)
Expand Down
Loading