openmpi 4.1.6/8 with ucx (1.18) broken
Open
@hppritcha is already working on this.
Since May 8, 2025.
bug
triage
- Dominant language
- Python
- Stars
- 5.1k
- Forks
- 2.5k
- Avg merge
- 3d 4h
- Merged PRs (30d)
- 82
Description
Steps to reproduce
spack install openmpi@:4 fabrics=cma,ucx
compilation is fine
Error message
when executing a mpi application one gets
symbol lookup error: /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/ucx-1.18.0-emtjpi3lgzhj5kooqe4onxod5wjk62iv/lib/ucx/libuct_rocm.so.0: undefined symbol: hsa_init
ldd gives :
ldd /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/ucx-1.18.0-emtjpi3lgzhj5kooqe4onxod5wjk62iv/lib/ucx/libuct_rocm.so
linux-vdso.so.1 (0x00007ffe715cb000)
libuct.so.0 => /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/ucx-1.18.0-emtjpi3lgzhj5kooqe4onxod5wjk62iv/lib/libuct.so.0 (0x00007f7f29f84000)
libucs.so.0 => /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/ucx-1.18.0-emtjpi3lgzhj5kooqe4onxod5wjk62iv/lib/libucs.so.0 (0x00007f7f29f0f000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f7f29ca2000)
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f7f29bb7000)
libucm.so.0 => /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/ucx-1.18.0-emtjpi3lgzhj5kooqe4onxod5wjk62iv/lib/libucm.so.0 (0x00007f7f29b97000)
/lib64/ld-linux-x86-64.so.2 (0x00007f7f29fe8000)
nm gives:
nm /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/ucx-1.18.0-emtjpi3lgzhj5kooqe4onxod5wjk62iv/lib/ucx/libuct_rocm.so | grep hsa
U hsa_agent_get_info
U hsa_amd_agent_iterate_memory_pools
U hsa_amd_agent_memory_pool_get_info
U hsa_amd_agents_allow_access
U hsa_amd_ipc_memory_attach
U hsa_amd_ipc_memory_create
U hsa_amd_ipc_memory_detach
U hsa_amd_memory_async_copy
U hsa_amd_memory_lock
U hsa_amd_memory_pool_allocate
U hsa_amd_memory_pool_free
U hsa_amd_memory_pool_get_info
U hsa_amd_memory_unlock
U hsa_amd_pointer_info
U hsa_init
U hsa_iterate_agents
U hsa_signal_create
U hsa_signal_destroy
U hsa_signal_load_scacquire
U hsa_signal_store_screlease
U hsa_signal_wait_scacquire
0000000000009110 t uct_rocm_hsa_agent_callback
0000000000009080 t uct_rocm_hsa_pool_callback
nm on libhsa-runtime gives
nm /nvme0/prudhomm/Devel/feelpp.spack/spack/opt/spack/__spack_path_placeholder__/__spack_path_placeholder__/__spack_path_placeholde/linux-x86_64_v3/hsa-rocr-dev-6.1.2-ulk6r53varidw3hwri4yttx3bi42dkhz/lib/libhsa-runtime64.so | grep hsa_init
this makes openmpi with ucx broken
Information on your system
- Spack: 1.0.0.dev0 (97edcb5acc364c8c61936d86ffb63c84c1d56af0)
- Python: 3.10.12
- Platform: linux-ubuntu22.04-zen2
General information
- I have run
spack debug reportand reported the version of Spack/Python/Platform - I have searched the issues of this repo and believe this is not a duplicate
- I have run the failing commands in debug mode and reported the output
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.