Add RIKYU support for SALMON - #75
Draft
william-dawson wants to merge 3 commits into
Draft
Conversation
Signed-off-by: William Dawson <william.dawson@riken.jp>
Signed-off-by: William Dawson <william.dawson@riken.jp>
william-dawson
marked this pull request as draft
July 31, 2026 06:59
Build: nvhpc/26.3 + OpenACC + cc100 (B200/Blackwell), with FORTRAN_COMPILER_HAS_MPI_VERSION3=OFF to work around the MPI_Iallreduce hang at >1 rank. Run: per-rank GPU binding via wrapper.sh (CUDA_VISIBLE_DEVICES), UCX_IB_GPU_DIRECT_RDMA=no for 2+ nodes, and runtime patching of the ¶llel section to nproc_ob=N (orbital-only decomposition, no real-space grid split) which is the verified GPU-optimal layout. Depends on RIKEN-RCCS#74 for RIKYU system/queue config (gpu partition, account=rkp00012). Signed-off-by: William Dawson <william.dawson@riken.jp> 💘 Generated with Crush Assisted-by: glm-5.2 via Crush <crush@charm.land>
william-dawson
force-pushed
the
add-salmon-rikyu
branch
from
July 31, 2026 07:12
f410203 to
bda79ba
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
cc100(B200/Blackwell), with-DFORTRAN_COMPILER_HAS_MPI_VERSION3=OFFto work around theMPI_Iallreducehang at >1 rank.wrapper.sh(CUDA_VISIBLE_DEVICES),UCX_IB_GPU_DIRECT_RDMA=nofor 2+ nodes, and runtime patching of the¶llelnamelist section tonproc_ob=N(orbital-only decomposition, no real-space grid split) — the verified GPU-optimal layout./data1/rkp00012/CX_input/SALMON/SALMON.tar.gz(overridable viaBK_SALMON_INPUT_ARCHIVE).Depends on #74 for RIKYU system/queue config (
gpupartition,account=rkp00012).Test plan
bash programs/salmon/build.sh RIKYUsucceeds on a RIKYU login/build nodebash programs/salmon/run.sh RIKYU 1 4 1produces correct GS+RT results (1 node, 4 GPU)bash programs/salmon/run.sh RIKYU 2 4 1produces correct results (2 nodes, 8 GPU, exercises UCX fix)¶llelsection is patched tonproc_ob=4(1-node) /nproc_ob=8(2-node) in the run logs💘 Generated with Crush