public_git
/
whisper.cpp
mirror of https://github.com/ggml-org/whisper.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

kleidiai : support for concurrent sme and neon kernel execution (llama/20070)

This commit is contained in:
Charles Xu 2026-03-10 08:25:25 +01:00 committed by Georgi Gerganov
parent bd64b8af4d
commit dfa6858d02
2 changed files with 987 additions and 274 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/ggml-cpu/kleidiai/kernels.cpp
View File

@ -520,7 +520,7 @@ static ggml_kleidiai_kernels gemm_gemv_kernels[] = {
/* .packed_stride_ex = */ &rhs_stride_fn4<kai_get_rhs_packed_stride_rhs_pack_nxk_qsi4c32pscalef16_qsu4c32s16s0>,
/* .pack_func_ex = */ &rhs_pack_fn12<kai_run_rhs_pack_nxk_qsi4c32pscalef16_qsu4c32s16s0>,
},
/* .required_cpu = */ CPU_FEATURE_DOTPROD | CPU_FEATURE_I8MM,
/* .required_cpu = */ CPU_FEATURE_I8MM,
/* .lhs_type = */ GGML_TYPE_F32,
/* .rhs_type = */ GGML_TYPE_Q4_0,
/* .op_type = */ GGML_TYPE_F32,
@ -631,7 +631,7 @@ static ggml_kleidiai_kernels gemm_gemv_kernels[] = {
/* .packed_stride_ex = */ &rhs_stride_fn4<kai_get_rhs_packed_stride_rhs_pack_nxk_qsi4c32pscalef16_qsu4c32s16s0>,
/* .pack_func_ex = */ &rhs_pack_fn12<kai_run_rhs_pack_nxk_qsi4c32pscalef16_qsu4c32s16s0>,
},
/* .required_cpu = */ CPU_FEATURE_DOTPROD | CPU_FEATURE_I8MM,
/* .required_cpu = */ CPU_FEATURE_I8MM,
/* .lhs_type = */ GGML_TYPE_F32,
/* .rhs_type = */ GGML_TYPE_Q4_0,
/* .op_type = */ GGML_TYPE_F32,
@ -801,7 +801,7 @@ static ggml_kleidiai_kernels gemm_gemv_kernels_q8[] = {
/* .packed_stride_ex = */ &rhs_stride_fn4<kai_get_rhs_packed_stride_rhs_pack_nxk_qsi8cxp_qsi8cx_neon>,
/* .pack_func_ex = */ &rhs_pack_scale_fn12<kai_run_rhs_pack_nxk_qsi8cxp_qsi8cx_neon>,
},
/* .required_cpu = */ CPU_FEATURE_DOTPROD | CPU_FEATURE_I8MM,
/* .required_cpu = */ CPU_FEATURE_I8MM,
/* .lhs_type = */ GGML_TYPE_F32,
/* .rhs_type = */ GGML_TYPE_Q8_0,
/* .op_type = */ GGML_TYPE_F32,

1255
ggml/src/ggml-cpu/kleidiai/kleidiai.cpp
View File

File diff suppressed because it is too large Load Diff