biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Feature][CPU Backend]: Optimize ARM vectorization backend (#30329)

Browse Source 
Signed-off-by: Radu Salavat <radu.salavat@arm.com>
This commit is contained in:
Radu Salavat
2026-02-03 04:17:56 +00:00
committed by GitHub
parent 5eac9a1b34
commit e69c990c21
5 changed files with 579 additions and 624 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
csrc/cpu/dnnl_kernels.cpp
View File

@@ -14,13 +14,11 @@ struct KernelVecType<float> {
using cvt_vec_type = vec_op::FP32Vec16;
};
#if !defined(__aarch64__) || defined(ARM_BF16_SUPPORT)
template <>
struct KernelVecType<c10::BFloat16> {
using load_vec_type = vec_op::BF16Vec16;
using cvt_vec_type = vec_op::FP32Vec16;
};
#endif
template <>
struct KernelVecType<c10::Half> {