File-based handshake - #5

Open

jychoi-hpc wants to merge 24 commits into

mainfrom

Commits on Aug 22, 2026

add file-based method
jychoi-hpc
committed
8f48c7d
add
jychoi-hpc
committed
b21a0db

Commits on Aug 24, 2026

Add FABRIC_IFACE selection from CPU affinity via hwloc
jychoi-hpc
committed
307cfb7
Add SLURM_STEP_NODELIST for DDP master_addr resolution
jychoi-hpc
committed
f40f182

Commits on Aug 26, 2026

Apply black formatting to Python files

jychoi-hpc
and
claude
committed
5403814
Fix multi-GPU DDP setup in VAE example

jychoi-hpc
and
claude
committed
923934f
Combine method=2 handshake into a single file per variable
jychoi-hpc
committed
8f25ac4
Fix multi-GPU DDP setup in VAE example
jychoi-hpc
committed
6b5c92d
Combine method=2 handshake into a single file per variable
jychoi-hpc
committed
6117437
Merge branch 'dev-file-handshake' of github.com:ORNL/DDStore into dev-file-handshake
jychoi-hpc
committed
976a217
fix setdevice
jychoi-hpc
committed
e50ecd1
Add cxi (Perlmutter) fabric support alongside hsn (Frontier), selected via DDSTORE_FABRIC_PROVIDER
jychoi-hpc
committed
5991499
Add --provider flag to cpu_nic_map.py CLI to preview cxi-translated NIC names
jychoi-hpc
committed
e680520
Remove -p/--pattern CLI flag from cpu_nic_map.py; kernel NIC names are always hsn* on both systems
jychoi-hpc
committed
bf84628
Simplify vae-ddp.py device selection to match vae_extra_train.py (unconditional cuda device, no per-rank set_device)
jychoi-hpc
committed
2f750e3
Set CUDA device by local-rank id only when multiple GPUs are visible, identically in vae-ddp.py and vae_extra_train.py
jychoi-hpc
committed
ce86fa8
Rename DDSTORE_FABRIC_PROVIDER to DDSTORE_FABRIC (hsn isn't a real libfabric provider name)
jychoi-hpc
committed
b26fdd6
Rename cpu_nic_map.py CLI flag --provider to --fabric, matching DDSTORE_FABRIC
jychoi-hpc
committed
40bc4ba

Commits on Aug 27, 2026

fix for aurora
Jong Youl Choi
committed
befa9ea
fix for aurora
Jong Youl Choi
committed
1e91ae4
remove
Jong Youl Choi
committed
3ea9ff1

Commits on Aug 28, 2026

Fix review comments: extern C linkage, localrank typo, README docs, path traversal, comm validation

Copilot
and
jychoi-hpc
authored
5a2089b
Potential fix for pull request finding

jychoi-hpc
and
Copilot
authored
e074d41
Potential fix for pull request finding

jychoi-hpc
and
Copilot
authored
d6263fa