Add VLLM_USE_SPINLOOP_EXT to use more efficient busy polling (#36517)

Signed-off-by: Patrick Schlangen <pschlan@amd.com>
This commit is contained in:
pschlan-amd
2026-05-11 16:11:49 -07:00
committed by GitHub
parent d7af6b34d8
commit 39dff5ff39
5 changed files with 259 additions and 14 deletions
+3
View File
@@ -686,6 +686,7 @@ class precompiled_wheel_utils:
"vllm/vllm_flash_attn/_vllm_fa2_C.abi3.so",
"vllm/vllm_flash_attn/_vllm_fa3_C.abi3.so",
"vllm/cumem_allocator.abi3.so",
"vllm/spinloop.abi3.so",
# ROCm-specific libraries
"vllm/_rocm_C.abi3.so",
]
@@ -993,6 +994,8 @@ if _is_cuda() or _is_hip():
# copying the relevant .py files from the source repository.
ext_modules.append(CMakeExtension(name="vllm.triton_kernels", optional=True))
ext_modules.append(CMakeExtension(name="vllm.spinloop"))
if _is_hip():
ext_modules.append(CMakeExtension(name="vllm._rocm_C"))