Skip to content

Fix more kernel launches over INT_MAX work-items with DPC++ 2026.2 #3104

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
antonwolfy wants to merge 2 commits into
base: master
Choose a base branch
from
+21 −4
Open

Fix more kernel launches over INT_MAX work-items with DPC++ 2026.2 #3104

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f19a532
Pass -fno-sycl-id-queries-fit-in-int to more SYCL modules
antonwolfy Oct 15, 2026
38a7201
Add a changelog entry
antonwolfy Oct 15, 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
1 change: 1 addition & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,6 +118,7 @@ This release is compatible with NumPy 2.5.
* Fixed incorrect results of `dpnp.tensor.vecdot` in some cases with strided outputs and of `dpnp.tensor` reductions, `dpnp.tensor.vecdot` and `dpnp.tensor.matmul` on large inputs with some data types [#3082](https://github.com/IntelPython/dpnp/pull/3082)
* Fixed `dpnp.median` and `dpnp.nanmedian` raising a `ValueError` for a tuple `axis` when a kept dimension has size 0 [#3081](https://github.com/IntelPython/dpnp/pull/3081)
* Fixed a `RuntimeError` raised by `dpnp.tensor` kernels launched over more than `INT_MAX` work-items when built with DPC++ compiler 2026.2 or newer, by also passing `-fno-sycl-id-queries-fit-in-int` to the linker [#3089](https://github.com/IntelPython/dpnp/pull/3089)
* Fixed a `RuntimeError` raised by kernels of the `ufunc`, `window` and `statistics` extensions and of the legacy backend launched over more than `INT_MAX` work-items when built with DPC++ compiler 2026.2 or newer [#3104](https://github.com/IntelPython/dpnp/pull/3104)

@vlad-perevezentsev vlad-perevezentsev Oct 19, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

there is a double space before the PR link


### Security

Expand Down
6 changes: 5 additions & 1 deletion dpnp/backend/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,7 +70,11 @@ endif()
# target_compile_definitions(${_trgt} PRIVATE _WIN=1)
# endif()

target_link_options(${_trgt} PUBLIC -fsycl-device-code-split=per_kernel)
target_compile_options(${_trgt} PUBLIC -fno-sycl-id-queries-fit-in-int)
target_link_options(
${_trgt}
PUBLIC -fsycl-device-code-split=per_kernel -fno-sycl-id-queries-fit-in-int
)

if(DPNP_GENERATE_COVERAGE)
target_link_options(${_trgt} PRIVATE -fprofile-instr-generate -fcoverage-mapping)
Expand Down
6 changes: 5 additions & 1 deletion dpnp/backend/extensions/statistics/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,7 +99,11 @@ else()
)
endif()

target_link_options(${python_module_name} PUBLIC -fsycl-device-code-split=per_kernel)
target_compile_options(${python_module_name} PUBLIC -fno-sycl-id-queries-fit-in-int)
target_link_options(
${python_module_name}
PUBLIC -fsycl-device-code-split=per_kernel -fno-sycl-id-queries-fit-in-int

@vlad-perevezentsev vlad-perevezentsev Oct 19, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

blas extension also needs this flag

@antonwolfy antonwolfy Oct 19, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

For what? It has no pure SYCL kernel.

@vlad-perevezentsev vlad-perevezentsev Oct 20, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

It does have one.
syrk.cpp has its own copy_kernel that fills the lower triangle after oneMKL syrk: on GPU it is cgh.parallel_for(sycl::range<2>{n, n}, ...) at syrk.cpp:116, and syrk_impl always calls it (syrk.cpp:228).
So a.T @ a with a = dpnp.ones((1, 46341), dtype="f4") launches it over 46341 * 46341 > INT_MAX work-items, the same case this PR fixes for ufunc/window/statistics.

)

if(DPNP_GENERATE_COVERAGE)
target_link_options(
Expand Down
6 changes: 5 additions & 1 deletion dpnp/backend/extensions/ufunc/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,7 +125,11 @@ else()
)
endif()

target_link_options(${python_module_name} PUBLIC -fsycl-device-code-split=per_kernel)
target_compile_options(${python_module_name} PUBLIC -fno-sycl-id-queries-fit-in-int)
target_link_options(
${python_module_name}
PUBLIC -fsycl-device-code-split=per_kernel -fno-sycl-id-queries-fit-in-int
)

if(DPNP_GENERATE_COVERAGE)
target_link_options(
Expand Down
6 changes: 5 additions & 1 deletion dpnp/backend/extensions/window/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,7 +94,11 @@ else()
)
endif()

target_link_options(${python_module_name} PUBLIC -fsycl-device-code-split=per_kernel)
target_compile_options(${python_module_name} PUBLIC -fno-sycl-id-queries-fit-in-int)
target_link_options(
${python_module_name}
PUBLIC -fsycl-device-code-split=per_kernel -fno-sycl-id-queries-fit-in-int
)

if(DPNP_GENERATE_COVERAGE)
target_link_options(
Expand Down
Loading