@Thyre Seems like this should be a generic easyblock instead, since it's not specific to one single software package, but will cover a whole group?

@Thyre Seems like this should be a generic easyblock instead, since it's not specific to one single software package, but will cover a whole group?

Yeah, I agree. I wasn't really sure initially where to place it, so I left it there for now.

I'll move it to the generic ones. As you've guessed, this can be used for most ROCm components.

Test report by @Thyre

Overview of tested easyconfigs (in order)

• SUCCESS HIP-6.4.1-rocm-compilers-19.0.0-ROCm-6.4.1.eb

Build succeeded for 1 out of 1 (total: 1 min 50 secs) (1 easyconfigs in total)
jrc0850.jureca - Linux Rocky Linux 9.6 (Blue Onyx), x86_64, AMD EPYC 7443 24-Core Processor (zen3), 8 x AMD AMD Instinct MI250X / MI250 (model: 0x740c, driver: "6.12.12"), Python 3.9.21
See https://gist.github.com/Thyre/fe11ea7428f1608d914ab8905d581983 for a full test report.

Test report by @Thyre

Overview of tested easyconfigs (in order)

• SUCCESS rocm-cmake-0.14.0-llvm-compilers-20.1.8.eb

• SUCCESS hipBLAS-common-1.4.0-llvm-compilers-20.1.8.eb

• SUCCESS hipBLAS-3.2.0-llvm-compilers-20.1.8-CUDA-12.9.1.eb

Build succeeded for 3 out of 3 (total: 54 secs) (3 easyconfigs in total)
jrc0900.jureca - Linux Rocky Linux 9.7, AArch64, ARM UNKNOWN (neoverse_v2), 1 x NVIDIA NVIDIA GH200 480GB, 590.48.01, Python 3.9.25
See https://gist.github.com/Thyre/fa6fca8e84810c3008293c280cb577ed for a full test report.

Test report by @Thyre

Overview of tested easyconfigs (in order)

• SUCCESS rocTracer-4.1.0-rocm-compilers-22.0.0-ROCm-7.2.0.eb

Build succeeded for 1 out of 1 (total: 45 secs) (1 easyconfigs in total)
Linux - Linux Arch Linux UNKNOWN, x86_64, AMD Ryzen 7 7800X3D 8-Core Processor (zen4), 1 x AMD Navi 48 [Radeon RX 9070/9070 XT/9070 GRE] (device id: 0x7550, gfx: gfx1201, driver: Linuxversion6.19.6-arch1-1(linux@archlinux)(gcc(GCC)15.2.120260209,GNUld(GNUBinutils)2.46)#1SMPPREEMPT_DYNAMICWed,04Mar202618:25:08+0000), 1 x AMD Raphael (device id: 0x164e, gfx: gfx1036, driver: Linuxversion6.19.6-arch1-1(linux@archlinux)(gcc(GCC)15.2.120260209,GNUld(GNUBinutils)2.46)#1SMPPREEMPT_DYNAMICWed,04Mar202618:25:08+0000), Python 3.14.3
See https://gist.github.com/Thyre/444798f64f9943f2fa4219bcf527e891 for a full test report.

Succesful test report using this easyblock here, for 10 builds that use this.

Old style EasyConfig with compiler_toolchain = 'rocm-llvm'

Test report by @Thyre

Overview of tested easyconfigs (in order)

• SUCCESS rocTracer-4.1.0-GCCcore-14.2.0-ROCm-6.4.1.eb

Build succeeded for 1 out of 1 (total: 1 min 36 secs) (1 easyconfigs in total)
jrc0850.jureca - Linux Rocky Linux 9.7 (Blue Onyx), x86_64, AMD EPYC 7443 24-Core Processor (zen3), 8 x AMD AMD INSTINCT MI250 (MCM) OAM AC MBA (device id: 0x740c, gfx: gfx90a, driver: 6.16.6), Python 3.9.25
See https://gist.github.com/Thyre/06e24934aa46acc57aa23e52c55c95d5 for a full test report.

The test correctly uses hipcc. The failure comes from rocTracer not supporting hipcc to build the code. That doesn't surprise me too much.

Test report by @Thyre

Overview of tested easyconfigs (in order)

• FAIL rocTracer-4.1.0-GCCcore-14.2.0-ROCm-6.4.1.eb (build issue)
  (partial log available at https://gist.github.com/Thyre/2be74c6a26ef824b42aea58dabc69524)

Build succeeded for 0 out of 1 (total: 39 secs) (1 easyconfigs in total)
jrc0850.jureca - Linux Rocky Linux 9.7 (Blue Onyx), x86_64, AMD EPYC 7443 24-Core Processor (zen3), 8 x AMD AMD INSTINCT MI250 (MCM) OAM AC MBA (device id: 0x740c, gfx: gfx90a, driver: 6.16.6), Python 3.9.25
See https://gist.github.com/Thyre/2d9b4ee401cbb80e55b930740014461a for a full test report.

This EasyBlock has a lot of succesful test reports for cross-compilation: 25 easyconfigs from easybuilders/easybuild-easyconfigs#25576 (comment) (with the one failing one being fixed and retested at easybuilders/easybuild-easyconfigs#25576 (comment)

A test build is currently running on jsc-zen3 easybuilders/easybuild-easyconfigs#25576 (comment) .

I'll do builds on a system with a native MI210 soon (hopefully tonight) using that same EasyConfigs PR. If that all succeds then, together with the test builds in which @Thyre already tested specific functionality, I think that's sufficient. This is all new functionality after all - it won't break anything existing for anyone. And I'd really like the unit tests on easybuilders/easybuild-easyconfigs#25576 to no longer complain about the missing easyblock, so we can clean up any remaining issues with unit tests there.