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 8, 16:30 UTC
fix(cpu): fix mla_decode compilation on x86 without AVX512 (#34052)
Signed-off-by: ihb2032 <hebome@foxmail.com> Co-authored-by: root <root@LAPTOP-FKNHV411.localdomain>
ihb2032 committed
Feb 9, 2026 at 16:55 UTC
5a5c43511ac98299856d0fee6c619fdd8bcdd2ef
1 file changed
+1
-10
csrc/cpu/mla_decode.cpp
+1
-10
@@ -38,16 +38,7 @@ struct KernelVecType<c10::BFloat16> {
38
using qk_vec_type = vec_op::BF16Vec32;
39
using v_load_vec_type = vec_op::BF16Vec16;
40
};
41
-
42
-#elif defined(__s390x__)
43
-template <>
44
-struct KernelVecType<c10::BFloat16> {
45
- using qk_load_vec_type = vec_op::BF16Vec16;
46
- using qk_vec_type = vec_op::FP32Vec16;
47
- using v_load_vec_type = vec_op::BF16Vec16;
48
-};
49
-
50
-#elif defined(__aarch64__)
41
+#else
42
template <>
43
struct KernelVecType<c10::BFloat16> {
44
using qk_load_vec_type = vec_op::BF16Vec16;