[DFT][MKLGPU] Update mklgpu support to mkl preparing for oneapi 2025.0 release #575
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Update how dft mklgpu backend include header files from MKL and how cmake add them to compilation. This is done to avoid name conflicts between interface and library backends. Force compiler to look for some header following the "" and <> inclusions rules accordingly with C++ Core Guideline SF.12
s-Nick
commented
Sep 27, 2024
| @@ -32,12 +32,14 @@ add_library(${LIB_OBJ} OBJECT | |||
| ) | |||
| add_dependencies(onemkl_backend_libs_dft ${LIB_NAME}) | |||
|
|
|||
| target_include_directories(${LIB_OBJ} | |||
| PUBLIC ${ONEMKL_INTERFACE_INCLUDE_DIRS} | |||
| ) | |||
There was a problem hiding this comment.
Giving the includes change I made a little bit down. This command is not necessary anymore, so I decided to remove it trying to keep the cmake as lean as possible.
Rbiessy
reviewed
Sep 27, 2024
Rbiessy
approved these changes
Oct 1, 2024
The new MKL release changed the message of unimplemented exception. Since tests were skipped by checking that message, some tests started to fail. Update skipping check to keep compatibility with both versions.
Removing enum elements that cause compilation failure in MKL 2025.0 Apparently also previous version don't need them.
|
Log using next MKL |
lhuot
approved these changes
Oct 9, 2024
| target_include_directories(${LIB_NAME} | ||
| PUBLIC ${ONEMKL_INTERFACE_INCLUDE_DIRS} | ||
| ) | ||
|
|
||
| # Due to using the same file name for different headers in this library and in | ||
| # MKL Intel library, we force the compiler to follow C++ Core Guideline SF.12 |
There was a problem hiding this comment.
Suggested change
| # MKL Intel library, we force the compiler to follow C++ Core Guideline SF.12 | |
| # the Intel(R) oneAPI Math Kernel Library, we force the compiler to follow C++ Core Guideline SF.12 |
Rbiessy
approved these changes
Oct 9, 2024
| @@ -75,7 +75,8 @@ class ComputeTests_real_real_out_of_place_REAL | |||
| } \ | |||
| catch (std::exception & e) { \ | |||
| std::string msg = e.what(); \ | |||
| if (msg.find("FFT_UNIMPLEMENTED") != std::string::npos) { \ | |||
| if ((msg.find("FFT_UNIMPLEMENTED") != std::string::npos) || \ | |||
| (msg.find("Unimplemented") != std::string::npos)) { \ | |||
There was a problem hiding this comment.
@s-Nick, was this addition motivated by a new exception that's not a oneapi::mkl::unimplemented exception with "Unimplemented" in its message? If yes, could you please share details?
acmnpv
pushed a commit
to gromacs/gromacs
that referenced
this pull request
Oct 21, 2024
Intel oneMKL deprecated the old descriptors in 2024.1 and is very vocal at runtime about it. Open-source oneMKL added the new descriptors in v0.5. As they don't have versioning macros just yet and we don't prioritize backward-compatibility there, we bump the minimal required version of oneMKL. Also update the include path for the upcoming oneMKL 2025.0 (ref: oneapi-src/oneMKL#575) and remove an outdated comment. Refs #4744
1 task
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Update how dft mklgpu backend includes header files from MKL and how cmake adds them to compilation.
This is done to avoid future name conflicts between this library and backend library. Forcing the compiler to look for some headers following the "" and <> inclusions rules accordingly with C++ Core Guideline SF.12
Checklist
All Submissions
Do all unit tests pass locally? Attach a log.
arc_mklgpu_dft_log.txt
N.B. Tests ran with current oneAPI release (2024.2) I cannot run tests with next one.
Have you formatted the code using clang-format?