Ecp-aurora-sdk-discuss

Download

[email protected]

March 2024

2 participants
2 discussions

Re: [Ecp-aurora-sdk-discuss] New oneAPI release and eng-compiler versions for sunspot.
by Bertoni, Colleen 29 Mar '24

29 Mar '24

Thanks Chris! Important note: with the new SDK drop, if you were specifying a std-c++ standard for OpenMP code, you’ll need to update it to -std=c++17 as a workaround or remove it (see below for more details). Comparing this new release to the previous default on Sunspot (oneapi/eng-compiler/2023.10.15.002) we see 44 fixes and 20 regressions in the bug reproducers (below). However, the 20 regressions fall in mainly 2 categories. Most of the regressions are from a compile-time bug in the C++ openmp compiler that complains that the compiler requires “-std=c++17” if the code was compiled with an older standard (-std=C++11 for example). The error message looks like: ``` static assertion failed due to requirement '201103L >= 201703L': DPCPP does not support C++ version earlier than C++17 ``` This can be worked around by changing it to “-std=c++17” for the time being. This is a bug, will not be required in the future, and already is fixed internally at Intel. Let us know if you have any questions, and thanks for all the bug reporting! Colleen Fixes: source/reproducers/dpcpp/ATSsubdevice_regression source/reproducers/dpcpp/VirtualFunction:VirtualFunction_minimal_ath_run [CMPLRLLVM-35295,XDEPS-6157] P0 source/reproducers/dpcpp/fortran_link:fortran_link_AOT [CMPLRLLVM-41268,cmplrllvm-44716] source/reproducers/dpcpp/kokkos_segfault_zebin [CMPLRLLVM-42875] source/reproducers/dpcpp/kokkos_sycl1a_bessel_fail_simd16 [GSD-6244] source/reproducers/dpcpp/kokkos_sycl2a_fp_model_precise_relwithdebinfo [GSD-3120] source/reproducers/dpcpp/syclmemcpy2d:syclmemcpy2d_ath_run [GSD-5515] P0 source/reproducers/dpcpp/work_item_larger_than_uint:work_item_larger_than_uint_max_work_item_no_fit-in-int [CMPLRLLVM-38306,XDEPS-4586] P0 source/reproducers/general/genxir_file source/reproducers/hybrid/oversubscribing:oversubscribing_more_device_memory_ath_run [GSD-786] source/reproducers/icx/ftrapv_error [CMPLRLLVM-44158] source/reproducers/ifx/complex_private:complex_private_WA [CMPLRLLVM-42381,GSD-3550,XDEPS-5672] P0 source/reproducers/ifx/emit_copy_from_ice [CMPLRLLVM-49292] source/reproducers/ifx/fopenmp-target-simd-data:fopenmp-target-simd-data_four source/reproducers/ifx/fopenmp-target-simd-data:fopenmp-target-simd-data_one source/reproducers/ifx/fopenmp-target-simd-data:fopenmp-target-simd-data_three source/reproducers/ifx/fopenmp-target-simd-data:fopenmp-target-simd-data_two source/reproducers/mkl/batched_dgemm_incorrect:batched_dgemm_incorrect_ath_compile [MKLD-16295] source/reproducers/mkl/batched_dgemm_incorrect:batched_dgemm_incorrect_batched_ath_run [MKLD-16295] source/reproducers/mkl/batched_dgemm_incorrect:batched_dgemm_incorrect_nobatched_ath_run [MKLD-16295] source/reproducers/mkl/minimal_zgeqrf_tester:minimal_zgeqrf_tester_bigger_ze00_ath_run [MKLD-15938] source/reproducers/mkl/minimal_zgeqrf_tester:minimal_zgeqrf_tester_ze00_ath_run [MKLD-15938] source/reproducers/mpi/openmp_target_zam:openmp_target_zam_zam1_ath_run [HPCS-8169] source/reproducers/opencl/concurrent_execution source/reproducers/openmp/FMM:FMM_ath_run [CMPLRLLVM-41769,CMPLRLLVM-41769(utils.F90)] source/reproducers/openmp/array_map:array_map_larger [GSD-5020,GSD-5208] source/reproducers/openmp/array_map:array_map_simple [GSD-5020,GSD-5208] source/reproducers/openmp/bigger_fmm_ice_array_reduction [CMPLRLLVM-48992] source/reproducers/openmp/bigger_fmm_ice_array_reduction:bigger_fmm_ice_array_reduction_n65 [CMPLRLLVM-48992] source/reproducers/openmp/gamess_mini_all_kernels:gamess_mini_all_kernels_j19 [CMPLRLLVM-45082,CMPLRLLVM-45890,CMPLRTST-21020,IGC-6246,XDEPS-5609,GSD-5514] source/reproducers/openmp/gamess_mini_all_kernels:gamess_mini_all_kernels_j21 [CMPLRLLVM-45082,CMPLRLLVM-45890,CMPLRTST-21020,IGC-6246,XDEPS-5609,GSD-5514] source/reproducers/openmp/gamess_mini_wrong_answer:gamess_mini_wrong_answer_offload_ath_compile [CMPLRLLVM-45082] source/reproducers/openmp/hanging_gamess:hanging_gamess_AOT_version [CMPLRLLVM-39824,GSD-2102,XDEPS-5609] source/reproducers/openmp/hanging_gamess:hanging_gamess_ath_run [CMPLRLLVM-39824,GSD-2102,XDEPS-5609] source/reproducers/openmp/helpers_thread_is_not_helping:helpers_thread_is_not_helping_2 [XDEPS-3493] P0 source/reproducers/openmp/helpers_thread_is_not_helping:helpers_thread_is_not_helping_threads_2 [XDEPS-3493] P0 source/reproducers/openmp/ifx-O2:ifx-O2_ath_run [CMPLRLLVM-43398] P0 source/reproducers/openmp/mem_transfer_hang:mem_transfer_hang_ath_run source/reproducers/openmp/memcopy_multidevice source/reproducers/openmp/multi_gpu_hanging:multi_gpu_hanging_parallel_for [GSD-2288] P0 source/reproducers/openmp/multi_gpu_ms67:multi_gpu_ms67_ath_run source/reproducers/openmp/qmcpack.cpu.qmcpack.ompsimd.sincos:qmcpack.cpu.qmcpack.ompsimd.sincos_ath_compile [CMPLRLLVM-24010,CMPLRLLVM-28600] source/reproducers/openmp/task_reduction [CMPLRLLVM-23590] source/reproducers/openmp/vla_quda:vla_quda_ath_run [GSD-4147,GSD-4255,GSD-5040,IGC-7270] P0 Regressions: > Related to omp needing std=c++17 bug: source/reproducers/openmp/snap source/reproducers/openmp/snap.collapse source/reproducers/openmp/su3_bench:su3_bench_ath_compile [CMPLRLLVM-10536] source/reproducers/openmp/su3_bench:su3_bench_ath_run [CMPLRLLVM-10536] source/reproducers/openmp/testsnap_ats source/reproducers/openmp/thermo:thermo_ath_compile [CMPLRLIBS-34271,CMPLRLLVM-52687] source/reproducers/openmp/thermo:thermo_ath_run [CMPLRLIBS-34271,CMPLRLLVM-52687] source/reproducers/openmp/thermo:thermo_static_library [CMPLRLIBS-34271,CMPLRLLVM-52687] source/reproducers/openmp/vpo [CMPLRLLVM-19583] source/reproducers/openmp/raja.kernel_lambda_args [CMPLRLLVM-11192] source/reproducers/openmp/miniqmc-smallrun [CMPLRLLVM-27865] P0 > Related to “omp simd” fails: source/reproducers/ifx/CMPLRLLVM-35621 source/reproducers/ifx/XDEPS-5141 source/reproducers/ifx/phasta_target_simd [CMPLRLLVM-35621,CMPLRLLVM-40195,CMPLRLLVM-52024,GSD-6634] P0 source/reproducers/openmp/simd_loops [CMPLRLLVM-38420] P0 > fail from using too much scratch, fixed in UMD 775 or UMD 803.7 source/reproducers/openmp/ddpp_gamess_mini_wrong_answer_fp_precise [CMPLRLLVM-45082] source/reproducers/openmp/gamess_mini_all_kernels:gamess_mini_all_kernels_j20 [CMPLRLLVM-45082,CMPLRLLVM-45890,CMPLRTST-21020,IGC-6246,XDEPS-5609,GSD-5514] source/reproducers/openmp/gamess_minispd_nan:gamess_minispd_nan_ath_compile [CMPLRLLVM-47316,GSD-2325,GSD-3402,GSD-4784,XDEPS-5607,XDEPS-6215] P0 > reported with new JIRAs: source/reproducers/dpcpp/supercontext:supercontext_shared source/reproducers/mkl/zgetrs_batch_slowdown [MKLD-15212,MKLD-15906,MKLD-16680] ________________________________ From: Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss-bounces(a)lists.jlse.anl.gov> on behalf of Chan-nui, Christopher via Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> Sent: Friday, March 29, 2024 11:27 AM To: ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> Subject: [Ecp-aurora-sdk-discuss] New oneAPI release and eng-compiler versions for sunspot. With the Sunspot update, a new version of the Aurora SDK is now available on Sunspot. The module files for this release, which are loaded by default: oneapi/eng-compiler/2023. 12. 15. 002 This is based on the 2024. 0 release of the oneAPI SDK. ZjQcmQRYFpfptBannerStart This Message Is From an External Sender This message came from outside your organization. ZjQcmQRYFpfptBannerEnd With the Sunspot update, a new version of the Aurora SDK is now available on Sunspot. The module files for this release, which are loaded by default: oneapi/eng-compiler/2023.12.15.002 This is based on the 2024.0 release of the oneAPI SDK. Release notes for the publicly available portions are available at: https://urldefense.us/v3/__https://www.intel.com/content/www/us/en/develope… <https://urldefense.us/v3/__https://www.intel.com/content/www/us/en/develope…> https://urldefense.us/v3/__https://www.intel.com/content/www/us/en/develope… <https://urldefense.us/v3/__https://www.intel.com/content/www/us/en/develope…> The engineering build contains bug fixes and performance improvements. Colleen will provide caveats, workarounds and other useful information about these releases shortly. -- Christopher Chan-Nui christopher.chan-nui(a)intel.com

1 0

New oneAPI release and eng-compiler versions for sunspot.
by Chan-nui, Christopher 29 Mar '24

29 Mar '24

With the Sunspot update, a new version of the Aurora SDK is now available on Sunspot. The module files for this release, which are loaded by default: oneapi/eng-compiler/2023.12.15.002 This is based on the 2024.0 release of the oneAPI SDK. Release notes for the publicly available portions are available at: https://urldefense.us/v3/__https://www.intel.com/content/www/us/en/develope… https://urldefense.us/v3/__https://www.intel.com/content/www/us/en/develope… The engineering build contains bug fixes and performance improvements. Colleen will provide caveats, workarounds and other useful information about these releases shortly. -- Christopher Chan-Nui christopher.chan-nui(a)intel.com

1 0