Skip to content

Fix khr_split_headers max_work_group_range test missing template args - #1248

Open
itlabs wants to merge 1 commit into
KhronosGroup:mainfrom
itlabs:fix-split-headers-max-wg-range
Open

itlabs wants to merge 1 commit into
KhronosGroup:mainfrom
itlabs:fix-split-headers-max-wg-range

Conversation

@itlabs

@itlabs itlabs commented Oct 11, 2026

Copy link
Copy Markdown

tests/extension/khr_split_headers/interop_max_work_group_queries.cpp (added in #1205) uses sycl::khr::info::device::max_work_group_range without template arguments. The sycl_khr_max_work_group_queries extension declares it as template <int Dimensions> struct max_work_group_range, so the test fails to compile:

error: use of class template 'sycl::khr::info::device::max_work_group_range' requires template arguments

Because all khr_split_headers sources are linked into one test_khr_split_headers executable, this breaks the entire category.

This PR checks max_work_group_range<1>, <2> and <3> instead, consistent with tests/extension/khr_max_work_group_queries.

Tested with DPC++ (-DSYCL_CTS_ENABLE_KHR_TESTS=ON) on Intel Data Center GPU Max 1100: all 461 test cases pass.

🤖 Generated with Claude Code

interop_max_work_group_queries.cpp named
sycl::khr::info::device::max_work_group_range without template arguments,
but the sycl_khr_max_work_group_queries extension declares it as
`template <int Dimensions> struct max_work_group_range`. The test therefore
fails to compile, and since all khr_split_headers sources are linked into a
single test_khr_split_headers executable, the whole category fails to build.

Check completeness of max_work_group_range<1>, <2> and <3> instead.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@itlabs
itlabs requested a review from a team as a code owner October 11, 2026 21:44
@CLAassistant

CLAassistant commented Oct 11, 2026 •

Copy link
Copy Markdown

CLA assistant check
All committers have signed the CLA.

@bader
bader requested a review from 0x12CC October 12, 2026 15:23
@TApplencourt

Copy link
Copy Markdown
Contributor

LGTM. Thanks a lot for taking the time to report and opening a PR @itlabs !

Now I'm curious how does the CTS even passed. My recollection is that we waited for protosycl implementing this KHR to merge the #1205

Digging...

ah protosycl doesn't implement sycl_khr_max_work_group_queries.

Mystery solved.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants