A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 00:30 UTC
[Hardware][Powerpc] Fix VLLM_CPU_OMP_THREADS_BIND="auto" low CPU utilization for Power (#27734)
Signed-off-by: Akash Kaothalkar <akash.kaothalkar@ibm.com> Co-authored-by: Akash Kaothalkar <akash.kaothalkar@ibm.com>
Akash kaothalkar committed
Oct 31, 2025 at 13:15 UTC
36960501d336a15cf0de7569e2662793ad9a4f3f
1 file changed
+2
-1
vllm/platforms/cpu.py
+2
-1
@@ -316,7 +316,8 @@ class CpuPlatform(Platform):
316
317
if (
318
platform.system() == "Linux"
319
- and Platform.get_cpu_architecture() == CpuArchEnum.ARM
319
+ and Platform.get_cpu_architecture()
320
+ in (CpuArchEnum.ARM, CpuArchEnum.POWERPC)
321
and not ("libomp" in ld_preload_str or "libgomp" in ld_preload_str)
322
):
323
# We need to LD_PRELOAD PyTorch's libgomp, otherwise only