Address oneAPI 2025 deprecations #278

yasahi-hpc · 2025-06-03T07:45:36Z

This PR addresses the oneAPI 2025 deprecations as found in Nightly builds.

tpadioleau

The changes are ok to me. I think there are remaining deprecation warnings like

/work/fft/src/KokkosFFT_SYCL_plans.hpp:126:51: warning: 'CONJUGATE_EVEN_STORAGE' is deprecated [-Wdeprecated-declarations]
  126 |   plan->set_value(oneapi::mkl::dft::config_param::CONJUGATE_EVEN_STORAGE,

/work/fft/src/KokkosFFT_SYCL_plans.hpp:112:9: warning: 'set_value' is deprecated: Use set_value(config_param, const std::vector<std::int64_t>&) instead. [-Wdeprecated-declarations]
  112 |   plan->set_value(oneapi::mkl::dft::config_param::INPUT_STRIDES,

science-enthusiast · 2025-06-03T11:41:51Z

fft/src/KokkosFFT_SYCL_plans.hpp

+  plan->set_value(oneapi::mkl::dft::config_param::FWD_STRIDES,
                  in_strides.data());
-  plan->set_value(oneapi::mkl::dft::config_param::OUTPUT_STRIDES,
+  plan->set_value(oneapi::mkl::dft::config_param::BWD_STRIDES,


It looks like the macros FWD_STRIDES and BWD_STRIDES are available in older versions of Intel MKL also. Have INPUT_STRIDES and OUTPUT_STRIDES alone been deprecated?

According to the errors, INPUT_STRIDES and OUTPUT_STRIDES have been deprecated.

FWD_STRIDES and INPUT_STRIDES are having the same meaning. Same with BWD_STRIDES and OUTPUT_STRIDES. I was curious about that.

yasahi-hpc · 2025-06-03T12:02:35Z

The changes are ok to me. I think there are remaining deprecation warnings like

/work/fft/src/KokkosFFT_SYCL_plans.hpp:126:51: warning: 'CONJUGATE_EVEN_STORAGE' is deprecated [-Wdeprecated-declarations]
  126 |   plan->set_value(oneapi::mkl::dft::config_param::CONJUGATE_EVEN_STORAGE,

/work/fft/src/KokkosFFT_SYCL_plans.hpp:112:9: warning: 'set_value' is deprecated: Use set_value(config_param, const std::vector<std::int64_t>&) instead. [-Wdeprecated-declarations]
 112 |   plan->set_value(oneapi::mkl::dft::config_param::INPUT_STRIDES,

Good catch.
I think the former is resolved.
The latter, we need to pass vectors directly, not pointers.

It may be that we can remove set_value(oneapi::mkl::dft::config_param::CONJUGATE_EVEN_STORAGE by using the default in the newer version

https://www.intel.com/content/www/us/en/docs/onemkl/developer-reference-dpcpp/2025-1/configuration-setting-member-functions.html#DFT-DESCRIPTOR-CLASS-TEMPLATE-CONFIGURATION-SETTING-FUNCTIONS

yasahi-hpc · 2025-06-04T08:01:31Z

@tpadioleau

Are you fine with the current change? Or you think it is better to suppress set_value(oneapi::mkl::dft::config_param::CONJUGATE_EVEN_STORAGE ?

Since I cannot test this on Intel GPU now, I would like to keep it until I confirm that it is OK to be removed

tpadioleau · 2025-06-04T08:29:20Z

Should we wait until we can test all the changes to merge then ?

tpadioleau · 2025-06-04T08:35:13Z

Has Kokkos fixed all these warnings about this_sub_group ?

yasahi-hpc · 2025-06-04T08:39:38Z

Has Kokkos fixed all these warnings about this_sub_group ?

Yes, with this
kokkos/kokkos#7918

In the nightly, you may find almost no warnings from kokkos codes

…eMKL

yasahi-hpc · 2025-10-01T18:47:11Z

People from Aurora and Intel told me the correct way to set up the descriptor with new arguments. Thanks a lot!
I have confirmed on Aurora that the code works correctly.

@tpadioleau @science-enthusiast
May I have your another review please

tpadioleau

That is good news but I cannot tell if we are addressing the warnings given the spurious warnings coming from Kokkos. Could we find a way to remove them ?

yasahi-hpc · 2025-10-01T19:30:15Z

That is good news but I cannot tell if we are addressing the warnings given the spurious warnings coming from Kokkos. Could we find a way to remove them ?

From nightly with newest Kokkos and OneAPI 2025, the warnings are from KokkosFFT side. All of these warnings should have been addressed in the PR.

/opt/intel/oneapi/mkl/2025.1/include/oneapi/mkl/dfti.hpp:18:2: warning: "Header 'oneapi/mkl/dfti.hpp' is deprecated. Please use oneapi/mkl/dft.hpp instead" [-W#warnings]
   18 | #warning "Header 'oneapi/mkl/dfti.hpp' is deprecated. Please use oneapi/mkl/dft.hpp instead"
      |  ^

/work/fft/src/KokkosFFT_SYCL_plans.hpp:24:15: warning: unused variable 'once' [-Wunused-variable]
   24 |   static bool once = [] {
      |               ^~~~
/work/fft/src/KokkosFFT_SYCL_plans.hpp:112:51: warning: 'INPUT_STRIDES' is deprecated: Use FWD/BWD strides instead. [-Wdeprecated-declarations]
  112 |   plan->set_value(oneapi::mkl::dft::config_param::INPUT_STRIDES,
      |                                                   ^
/opt/intel/oneapi/mkl/2025.1/include/oneapi/mkl/dft.hpp:51:31: note: 'INPUT_STRIDES' has been explicitly marked deprecated here
   51 |     INPUT_STRIDES           [[deprecated("Use FWD/BWD strides instead.")]]
      |                               ^

/work/fft/src/KokkosFFT_SYCL_plans.hpp:114:51: warning: 'OUTPUT_STRIDES' is deprecated: Use FWD/BWD strides instead. [-Wdeprecated-declarations]
  114 |   plan->set_value(oneapi::mkl::dft::config_param::OUTPUT_STRIDES,
      |                                                   ^
/opt/intel/oneapi/mkl/2025.1/include/oneapi/mkl/dft.hpp:53:31: note: 'OUTPUT_STRIDES' has been explicitly marked deprecated here
   53 |     OUTPUT_STRIDES          [[deprecated("Use FWD/BWD strides instead.")]]
      |                               ^

/work/fft/src/KokkosFFT_SYCL_plans.hpp:126:51: warning: 'CONJUGATE_EVEN_STORAGE' is deprecated [-Wdeprecated-declarations]
  126 |   plan->set_value(oneapi::mkl::dft::config_param::CONJUGATE_EVEN_STORAGE,
      |                                                   ^
/opt/intel/oneapi/mkl/2025.1/include/oneapi/mkl/dft.hpp:49:31: note: 'CONJUGATE_EVEN_STORAGE' has been explicitly marked deprecated here
   49 |     CONJUGATE_EVEN_STORAGE  [[deprecated]]  = DFTI_CONJUGATE_EVEN_STORAGE,
      |                               ^

tpadioleau

In some parts of the PR, we are changing the implementation for versions <20250100. If we allow this, can't we only keep the version >=20250100 ? Otherwise maybe keep the old implementation for <20250100 ?

yasahi-hpc · 2025-10-02T11:48:51Z

In some parts of the PR, we are changing the implementation for versions <20250100. If we allow this, can't we only keep the version >=20250100 ? Otherwise maybe keep the old implementation for <20250100 ?

That is a good point.
The problem is that I am not sure whether the version >=20250100 is functional for our requirement of IntelLLVM 2023.0.0+.
As soon as we make the minimum compiler to IntelLLVM 2025.0.0+, we can safely drop the older implementation.

tpadioleau · 2025-10-02T12:17:28Z

In some parts of the PR, we are changing the implementation for versions <20250100. If we allow this, can't we only keep the version >=20250100 ? Otherwise maybe keep the old implementation for <20250100 ?

That is a good point. The problem is that I am not sure whether the version >=20250100 is functional for our requirement of IntelLLVM 2023.0.0+. As soon as we make the minimum compiler to IntelLLVM 2025.0.0+, we can safely drop the older implementation.

Then I would consider keeping the safe old implementation when <20250100.

fft/src/KokkosFFT_SYCL_plans.hpp

yasahi-hpc · 2025-10-02T12:25:09Z

In some parts of the PR, we are changing the implementation for versions <20250100. If we allow this, can't we only keep the version >=20250100 ? Otherwise maybe keep the old implementation for <20250100 ?

That is a good point. The problem is that I am not sure whether the version >=20250100 is functional for our requirement of IntelLLVM 2023.0.0+. As soon as we make the minimum compiler to IntelLLVM 2025.0.0+, we can safely drop the older implementation.

Then I would consider keeping the safe old implementation when <20250100.

Do you mean that the default should be the new version like

#if INTEL_MKL_VERSION < 20250100
#include <oneapi/mkl/dfti.hpp>
#else
#include <oneapi/mkl/dft.hpp>
#endif

fft/src/KokkosFFT_SYCL_plans.hpp

yasahi-hpc · 2025-10-02T13:07:04Z

@tpadioleau I have fixed accordingly. May I have another review?

fft/src/KokkosFFT_SYCL_plans.hpp

tpadioleau

LGTM, thanks

yasahi-hpc · 2025-10-02T15:47:13Z

Thanks! I will merge this.
Hope all the warnings are suppressed in nightly

yasahi-hpc · 2025-10-03T11:54:02Z

@tpadioleau

Nightly looks almost good.

https://github.com/kokkos/kokkos-fft/actions/runs/18210223243/job/51849125743

I do not know how to suppress

These options as listed are not supported.
For more information, use '-qnextgen-diag'.
option list: 
	-Winline

tpadioleau · 2025-10-03T11:57:03Z

@tpadioleau

Nightly looks almost good.

https://github.com/kokkos/kokkos-fft/actions/runs/18210223243/job/51849125743

I do not know how to suppress
These options as listed are not supported.
For more information, use '-qnextgen-diag'.
option list: 
	-Winline

Can you check that this flag is being added by googletest ?

yasahi-hpc · 2025-10-03T12:08:40Z

@tpadioleau
Nightly looks almost good.
https://github.com/kokkos/kokkos-fft/actions/runs/18210223243/job/51849125743
I do not know how to suppress
These options as listed are not supported.
For more information, use '-qnextgen-diag'.
option list: 
	-Winline
Can you check that this flag is being added by googletest ?

How can I check that?

tpadioleau · 2025-10-03T12:21:10Z

@tpadioleau
Nightly looks almost good.
https://github.com/kokkos/kokkos-fft/actions/runs/18210223243/job/51849125743
I do not know how to suppress
These options as listed are not supported.
For more information, use '-qnextgen-diag'.
option list: 
	-Winline
Can you check that this flag is being added by googletest ?
How can I check that?

You can build in verbose mode

cmake --build ... --verbose

yasahi-hpc · 2025-10-03T12:26:54Z

@tpadioleau
Nightly looks almost good.
https://github.com/kokkos/kokkos-fft/actions/runs/18210223243/job/51849125743
I do not know how to suppress
These options as listed are not supported.
For more information, use '-qnextgen-diag'.
option list: 
	-Winline
Can you check that this flag is being added by googletest ?
How can I check that?
You can build in verbose mode

cmake --build ... --verbose

Can I test this in nightly?
On Aurora, I cannot built with oneMKL 2025+.

The oneAPI is 2025+ but oneMKL is 24.347.

yasahi-hpc self-assigned this Jun 3, 2025

yasahi-hpc added the cleanup label Jun 3, 2025

tpadioleau requested changes Jun 3, 2025

View reviewed changes

science-enthusiast reviewed Jun 3, 2025

View reviewed changes

Yuuichi Asahi added 4 commits October 2, 2025 02:58

Address oneAPI 2025 deprecations

a867709

remove duplicated comments

f053b9f

Upass vectors rather than pointers to set_value function for newer on…

4d10914

…eMKL

fix: FWD_STRIDES and BWD_STRIDES in SYCL plan creation

642696a

yasahi-hpc force-pushed the address-oneAPI-2025-deprecations branch from fcd62a7 to 642696a Compare October 1, 2025 18:00

tpadioleau reviewed Oct 1, 2025

View reviewed changes

tpadioleau reviewed Oct 2, 2025

View reviewed changes

fft/src/KokkosFFT_SYCL_plans.hpp Outdated Show resolved Hide resolved

tpadioleau reviewed Oct 2, 2025

View reviewed changes

fft/src/KokkosFFT_SYCL_plans.hpp Outdated Show resolved Hide resolved

use older notation for INTEL_MKL_VERSION < 20250100

cb2ce65

tpadioleau reviewed Oct 2, 2025

View reviewed changes

fft/src/KokkosFFT_SYCL_plans.hpp Outdated Show resolved Hide resolved

use INPUT_STRIDES/OUTPUT_STRIDES to support oneAPI < 2024.2

1284cca

tpadioleau approved these changes Oct 2, 2025

View reviewed changes

yasahi-hpc merged commit b15fce5 into kokkos:main Oct 2, 2025
25 checks passed

yasahi-hpc deleted the address-oneAPI-2025-deprecations branch October 2, 2025 15:47

Address oneAPI 2025 deprecations #278

Address oneAPI 2025 deprecations #278

Uh oh!

Conversation

yasahi-hpc commented Jun 3, 2025

Uh oh!

tpadioleau left a comment

Choose a reason for hiding this comment

Uh oh!

science-enthusiast Jun 3, 2025

Choose a reason for hiding this comment

Uh oh!

yasahi-hpc Jun 3, 2025

Choose a reason for hiding this comment

Uh oh!

science-enthusiast Jun 3, 2025

Choose a reason for hiding this comment

Uh oh!

yasahi-hpc commented Jun 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yasahi-hpc commented Jun 4, 2025

Uh oh!

tpadioleau commented Jun 4, 2025

Uh oh!

tpadioleau commented Jun 4, 2025

Uh oh!

yasahi-hpc commented Jun 4, 2025

Uh oh!

yasahi-hpc commented Oct 1, 2025

Uh oh!

tpadioleau left a comment

Choose a reason for hiding this comment

Uh oh!

yasahi-hpc commented Oct 1, 2025

Uh oh!

tpadioleau left a comment

Choose a reason for hiding this comment

Uh oh!

yasahi-hpc commented Oct 2, 2025

Uh oh!

tpadioleau commented Oct 2, 2025

Uh oh!

Uh oh!

yasahi-hpc commented Oct 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

yasahi-hpc commented Oct 2, 2025

Uh oh!

Uh oh!

tpadioleau left a comment

Choose a reason for hiding this comment

Uh oh!

yasahi-hpc commented Oct 2, 2025

Uh oh!

Uh oh!

yasahi-hpc commented Oct 3, 2025

Uh oh!

tpadioleau commented Oct 3, 2025

Uh oh!

yasahi-hpc commented Oct 3, 2025

Uh oh!

tpadioleau commented Oct 3, 2025

Uh oh!

yasahi-hpc commented Oct 3, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

yasahi-hpc commented Jun 3, 2025 •

edited

Loading

yasahi-hpc commented Oct 2, 2025 •

edited

Loading