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 5, 11:00 UTC
docs: clarify SMT and OMP acronyms in CpuPlatform (#39085)
Md. Mekayel Anik committed
Apr 8, 2026 at 06:42 UTC
d521dcdbcce68f7f87fceccde5223c57e9b66e70
1 file changed
+3
-1
vllm/platforms/cpu.py
+3
-1
@@ -76,7 +76,9 @@ class CpuPlatform(Platform):
76
dist_backend: str = "gloo"
77
device_control_env_var = "CPU_VISIBLE_MEMORY_NODES"
78
omp_process_manager = None
79
- smt = 1 # SMT level for OMP - 4 threads on PowerPC, 1 on others
79
+ # Simultaneous Multithreading (SMT) level for OpenMP:
80
+ # 4 on PowerPC, 1 on non-PowerPC architectures
81
+ smt = 1
82
global_cpu_mask = None
83
simulate_numa = int(os.environ.get("_SIM_MULTI_NUMA", 0))
84