[ROCm] Implement GPU-to-NUMA-node detection (#40015)

Signed-off-by: Patrick Schlangen <pschlan@amd.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
This commit is contained in:
pschlan-amd
2026-04-23 10:08:48 -05:00
committed by GitHub
co-authored by TJian
parent 53ecc807c0
commit 0098db9ec1
2 changed files with 31 additions and 3 deletions
+3 -3
View File
@@ -155,9 +155,9 @@ switch to `--physcpubind=<cpu-list> --membind=<node>`.
These `--numa-bind*` options only apply to GPU execution processes. They do not
configure the CPU backend's separate thread-affinity controls. Automatic
GPU-to-NUMA detection is currently implemented for CUDA/NVML-based platforms;
other GPU backends must provide explicit binding lists if they use these
options.
GPU-to-NUMA detection is currently implemented for CUDA/NVML-based as well as
ROCM-based platforms; other GPU backends must provide explicit binding lists if
they use these options.
`--numa-bind-nodes` takes one non-negative NUMA node index per visible GPU, in
the same order as the GPU indices.