Skip to content
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

[tests][cmake] an update of ONEDPL_BACKEND variable description #1953

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[tests][cmake] an update of ONEDPL_BACKEND variable description #1953

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
2 changes: 1 addition & 1 deletion cmake/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ The following variables are provided for oneDPL configuration:

| Variable | Type | Description | Default value |
|------------------------------|--------|-----------------------------------------------------------------------------------------------|---------------|
| ONEDPL_BACKEND | STRING | Threading backend; supported values: tbb, dpcpp, dpcpp_only, serial, ...; the default value is defined by compiler: dpcpp for DPC++ and tbb for others | tbb/dpcpp |
| ONEDPL_BACKEND | STRING | Threading backend; supported values: tbb, dpcpp, dpcpp_only, tbb, omp, serial, ...; the default value is defined by compiler: dpcpp for DPC++ and tbb for others. These values may conflict with explicit set variables like `-DCMAKE_CXX_FLAGS="-DONEDPL_USE_DPCPP_BACKEND=0"` etc. | tbb/dpcpp |
Copy link
Contributor

@dmitriy-sobolev dmitriy-sobolev Dec 2, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

Suggested change
| ONEDPL_BACKEND | STRING | Threading backend; supported values: tbb, dpcpp, dpcpp_only, tbb, omp, serial, ...; the default value is defined by compiler: dpcpp for DPC++ and tbb for others. These values may conflict with explicit set variables like `-DCMAKE_CXX_FLAGS="-DONEDPL_USE_DPCPP_BACKEND=0"` etc. | tbb/dpcpp |
| ONEDPL_BACKEND | STRING | Threading backend; supported values: `dpcpp`, `dpcpp_only`, `tbb`, `omp`, `serial`. The default value is `dpcpp` if SYCL support can be detected and `tbb` otherwise. Avoid defining backend control macros (`ONEDPL_USE_TBB_BACKEND`, `ONEDPL_USE_OPENMP_BACKEND`, `ONEDPL_USE_DPCPP_BACKEND`) directly, as they may conflict with the `ONEDPL_BACKEND` setting. | tbb/dpcpp |

Copy link
Contributor

@dmitriy-sobolev dmitriy-sobolev Dec 2, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

The suggestion above also replaces

the default value is defined by compiler: dpcpp for DPC++ and tbb for others.

with

The default value is dpcpp if SYCL support can be detected and tbb otherwise.

It aligns the note with the relevant behaviour.

Copy link
Contributor

Choose a reason for hiding this comment

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

Avoid defining backend control macros (ONEDPL_USE_TBB_BACKEND, ONEDPL_USE_OPENMP_BACKEND, ONEDPL_USE_DPCPP_BACKEND) directly

Would not it be better to describe what takes precedence in this case? Or is it not deterministic?

Copy link
Contributor

@dmitriy-sobolev dmitriy-sobolev Dec 6, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

It is deterministic. See the mapping table below.

ONEDPL_BACKEND cmake option oneDPL backend macro
dpcpp (default if SYCL is available) -
dpcpp_only ONEDPL_USE_TBB_BACKEND=0
tbb (default if SYCL is not available) ONEDPL_USE_DPCPP_BACKEND=0
omp ONEDPL_USE_DPCPP_BACKEND=0; ONEDPL_USE_OPENMP_BACKEND=1; ONEDPL_USE_TBB_BACKEND=0
serial ONEDPL_USE_DPCPP_BACKEND=0; ONEDPL_USE_OPENMP_BACKEND=0; ONEDPL_USE_TBB_BACKEND=0

I struggle to find a common logic between the options for a crisp description...

This is orthogonal to the question above, but I would prefer this set of options:

cmake flag and its options oneDPL backend macro
ONEDPL_HOST_PARALLEL_BACKEND
auto (default) (select the backend basing on the internal header heuristics)
tbb ONEDPL_USE_TBB_BACKEND=1
openmp ONEDPL_USE_OPENMP_BACKEND=1
serial ONEDPL_USE_OPENMP_BACKEND=0; ONEDPL_USE_TBB_BACKEND=0
ONEDPL_DEVICE_BACKEND
auto (default) (select the backend basing on the internal header heuristics)
dpcpp ONEDPL_USE_DPCPP_BACKEND=1
off ONEDPL_USE_DPCPP_BACKEND=0

I am going to create an issue with a feature request.

| ONEDPL_DEVICE_TYPE | STRING | Select device type for oneDPL test targets; affects only DPC++ backends; supported values: GPU, CPU, FPGA_HW, FPGA_EMU | GPU |
| ONEDPL_DEVICE_BACKEND | STRING | Select device backend type for oneDPL test targets; affects only oneDPL DPC++ backends; supported values: opencl, level_zero, cuda, hip or * (the best backend as per DPC++ runtime heuristics). | * |
| ONEDPL_USE_UNNAMED_LAMBDA | BOOL | Pass `-fsycl-unnamed-lambda`, `-fno-sycl-unnamed-lambda` compile options or nothing | |
Expand Down
Loading