Update alpaka to 2.1.1 "Fixed Transformation" [16.0.x]#10260
Conversation
This release implements a new interface, similar to std::transform, that simplifies writing asynchronous parallel algorithms across all back-ends. SYCL support is extended to NVIDIA and AMD GPUs. The release introduces unified memory and expands asynchronous memory allocation to buffers of any dimension. Interoperability with standard C++ is improved through std::span support: alpaka buffers expose a span interface, and any std::span can be used as an alpaka view. It adds compile-time warp-size definitions, extends atomic increment and decrement operations and fixes their behaviour on CPU back-end; it introduces a C++ concept for alpaka accelerators together with new type traits, along with many smaller fixes and improvements. The CI has been updated to test newer operating systems and compilers, including Clang 20 and ROCm 6.3, 6.4, and 7.0. The full list of changes is available in the ChangeLog.
|
enable gpu |
|
please test |
|
A new Pull Request was created by @fwyzard for branch IB/CMSSW_16_0_X/master. @akritkbehera, @iarspider, @raoatifshad, @smuzaffar can you please review it and eventually sign? Thanks.
|
|
cms-bot internal usage |
|
backport #10250 |
|
+1 Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-927186/50364/summary.html Comparison SummarySummary:
AMD_MI300X Comparison SummarySummary:
AMD_W7900 Comparison SummarySummary:
NVIDIA_H100 Comparison SummaryThere are some workflows for which there are errors in the baseline: Summary:
NVIDIA_L40S Comparison SummarySummary:
NVIDIA_T4 Comparison SummarySummary:
|
|
@gartung , can you please check the max memory comparison job. It failed for h100 gpu . the link for max memory failed job is broken at https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-927186/50364/summary.html . It points to missing https://cmssdt.cern.ch/SDT/jenkins-artifacts/baseLineComparisonsNVIDIA_H100/CMSSW_16_0_X_2025-12-18-2300+927186/72479/maxmem-comparison/maxmem_summary.html file. In the log file https://cmssdt.cern.ch/SDT/jenkins-artifacts/baseLineComparisonsNVIDIA_H100/CMSSW_16_0_X_2025-12-18-2300+927186/72479/maxmem-comparison/maxmem_summary.log I see messages like |
|
@fwyzard , is this ready to go in or do you want to run more local tests? |
|
ah this is for 16.0.X, lets wait for this to be integrated in 16.1.X first |
|
Both 16.1.x and 16.0.x PRs are good to go for me. |
|
REMINDER @mandrenguyen, @ftenchini, @sextonkennedy: This PR was tested with cms-sw/cmssw#49848, please check if they should be merged together |
|
+externals @cms-sw/orp-l2 feel free to include it for next 16.0.X IB/release. Note that, to avoid compilation warnings, we also need cms-sw/cmssw#49848 to go with this change |
cmsbuild
added
externals-approved
fully-signed
and removed
externals-pending
pending-signatures
labels
|
This pull request is fully signed and it will be integrated in one of the next IB/CMSSW_16_0_X/master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @mandrenguyen, @sextonkennedy, @ftenchini (and backports should be raised in the release meeting by the corresponding L2) |
|
+1 |
cmsbuild
merged commit 4ed7b22
into
cms-sw:IB/CMSSW_16_0_X/master
4 participants
This release implements a new interface, similar to std::transform, that simplifies writing asynchronous parallel algorithms across all back-ends. SYCL support is extended to NVIDIA and AMD GPUs.
The release introduces unified memory and expands asynchronous memory allocation to buffers of any dimension. Interoperability with standard C++ is improved through std::span support: alpaka buffers expose a span interface, and any std::span can be used as an alpaka view. It adds compile-time warp-size definitions, extends atomic increment and decrement operations and fixes their behaviour on CPU back-end; it introduces a C++ concept for alpaka accelerators together with new type traits, along with many smaller fixes and improvements. The CI has been updated to test newer operating systems and compilers, including Clang 20 and ROCm 6.3, 6.4, and 7.0.
The full list of changes is available in the ChangeLog.