mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 34 Packages Projects Releases Wiki Activity

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

Browse Source
This commit is contained in:
Charles Xu
2026-03-10 08:25:25 +01:00
committed by GitHub
parent af237f3026
commit 0cd4f4720b
3 changed files with 994 additions and 275 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cpu/kleidiai/kernels.cpp
+3 -3
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,
ggml/src/ggml-cpu/kleidiai/kleidiai.cpp
+984 -271
View File
File diff suppressed because it is too large Load Diff