iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[CPU] Add AMX-only high-performance MLA backend for DeepSeek V2/V3/R1 (#52616)

Browse Source 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
This commit is contained in:
Li, Jiang
2026-08-19 10:28:52 +08:00
committed by GitHub
parent f485081e8b
commit e4d61d0d22
20 changed files with 4156 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
cmake/cpu_extension.cmake
+5 -1
View File
@@ -519,7 +519,11 @@ if (ENABLE_X86_ISA)
"csrc/cpu/sgl-kernels/moe.cpp"
"csrc/cpu/sgl-kernels/moe_int8.cpp"
"csrc/cpu/sgl-kernels/moe_int4.cpp"
"csrc/cpu/sgl-kernels/moe_fp8.cpp")
"csrc/cpu/sgl-kernels/moe_fp8.cpp"
"csrc/cpu/sgl-kernels/bmm.cpp"
"csrc/cpu/sgl-kernels/decode.cpp"
"csrc/cpu/sgl-kernels/extend.cpp"
"csrc/cpu/sgl-kernels/mla_cache.cpp")
set(VLLM_EXT_SRC_AVX512
"csrc/cpu/sgl-kernels/fla.cpp"