Source: petsc Version: 3.20.6+dfsg1-3 Severity: wishlist X-Debbugs-Cc: c...@slerp.xyz
Dear Maintainer, The petsc toolkit has an AMD ROCm backend for hardware acceleration on AMD GPUs. It would be good to enable this functionality. The AMD backend depends on the following packages: hipcc libamdhip64-dev libhipblas-dev libhiprand-dev libhipsparse-dev libhipsolver-dev librocsparse-dev librocsolver-dev librocblas-dev librocrand-dev librocthrust-dev librocprim-dev The petsc toolkit may require kokkos with ROCm support, which has been requested in Debian Bug #1082813 [1]. The ROCm libraries packaged for Debian are currently built for gfx803, gfx900, gfx906, gfx908, gfx90a, gfx1010, gfx1030, gfx1100, gfx1101, gfx1102. They will soon be adding gfx942. However, kokkos appears to be limited to gfx900, gfx906, gfx908, gfx90a, gfx940, gfx1030, and gfx1100. I supppose petsc may be limited to the kokkos-supported architectures. In controlling the ROCm portion of the petsc build, relevent configuration flags include --with-hip-arch=<arch>, HIPPPFLAGS=<flags>, and --with-hip-lib=<libs>. I would suggest adding -DROCPRIM_USE_ARCH_CONVERSION to the DEB_CXXFLAGS_MAINT_PREPEND flags when building petsc. This is a rocprim build option that affects rocthrust. It causes gfx902, gfx909, and gfx90c to use gfx900 code paths and causes gfx1031, gfx1032, gfx1033, gfx1034, gfx1035, and gfx1036 to use gfx1030 code paths. It may or may not be sufficient for petsc support to those architectures, but it will at least be necessary. If you require access to compatible hardware to test this functionality, please reach out privately and I may be able to help. Sincerely, Cory Bloor [1]: https://bugs.debian.org/1082813 -- System Information: Debian Release: trixie/sid APT prefers unstable APT policy: (500, 'unstable') Architecture: amd64 (x86_64) Kernel: Linux 6.10.11-amd64 (SMP w/32 CPU threads; PREEMPT) Locale: LANG=C, LC_CTYPE=C.UTF-8 (charmap=UTF-8), LANGUAGE not set Shell: /bin/sh linked to /usr/bin/dash Init: unable to detect
