|
| 1 | +--- |
| 2 | +authors: [toine, timvc, boegel, ocaisa, darkless] |
| 3 | +date: 2025-06-24 |
| 4 | +slug: rocm |
| 5 | +--- |
| 6 | + |
| 7 | +# Building ROCm Support in EESSI |
| 8 | + |
| 9 | +* an overview of what we're doing (adding ROCm support, previous blog post, etc) |
| 10 | + |
| 11 | +<!-- more --> |
| 12 | + |
| 13 | +## Our Goals: The ROCm Integration Roadmap |
| 14 | + |
| 15 | +* main objective: full rocm support in EESSI |
| 16 | +* milestones: |
| 17 | + * building core ROCm components (in EESSI) |
| 18 | + * driver support |
| 19 | + * validation suite / examples running |
| 20 | + |
| 21 | +## Foundation Work: What We Did Before |
| 22 | + |
| 23 | +* ecosystem mapping: ROCm overview (previous blog post) |
| 24 | +* used these as a starting point (started on version 6.3.3) |
| 25 | + * https://github.com/bedroge/eb-rocm/tree/main (Bob) |
| 26 | + * https://github.com/Thyre/easybuild-custom/tree/support-passing-amdgcn/easybuild/easyconfigs/r (Jan) |
| 27 | +* Davide got ROCm-LLVM to build with EasyBuild |
| 28 | + * https://github.com/easybuilders/easybuild-easyblocks/pull/3706 |
| 29 | + * https://github.com/easybuilders/easybuild-easyblocks/pull/3781 |
| 30 | + |
| 31 | +## Current Progress: Building the Core Stack |
| 32 | + |
| 33 | +* for version 6.4.0 the following can build with EasyBuild in the EESSI build container |
| 34 | + * ROCm-LLVM |
| 35 | + * rocminfo |
| 36 | + * rocm-cmake |
| 37 | + * HIP |
| 38 | + * amdsmi |
| 39 | + * roctracer |
| 40 | +* currently testing/building on a VM without AMD GPU |
| 41 | + * some sanity checks are skipped |
| 42 | + |
| 43 | +details: |
| 44 | + |
| 45 | +* works in an EESSI build container/environment |
| 46 | +* EasyBuild recipes created and tested |
| 47 | +* some patches / workarounds / hooks used |
| 48 | + |
| 49 | +limitations: |
| 50 | + |
| 51 | +* no runtime validation yet |
| 52 | +* skipped sanity checks (no GPU) |
| 53 | + |
| 54 | +## Next Steps: Validating the Integration |
| 55 | + |
| 56 | +high priority: |
| 57 | + |
| 58 | +* add support for the driver/GPU |
| 59 | +* finish adding support for the core components and validation suite / examples in EasyBuild |
| 60 | +* run validation suite / examples to check that everything actually works |
| 61 | +* test building with sanity checks enabled |
| 62 | + |
| 63 | +lower priority: |
| 64 | + |
| 65 | +* add support for more libraries and frameworks |
| 66 | +* add support for popular scientific applications |
| 67 | +* update ROCm overview with experience from building / running ROCm |
| 68 | +* see about contributing overview back to AMD docs |
0 commit comments