tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-28 20:25:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml-cpu: Add CPU backend support for KleidiAI library (#11390)

Browse Source 
* ggml-cpu: Add CPU backend support for KleidiAI library

* Add environmental variable GGML_KLEIDIAI_SME

* Add support for multithread LHS conversion

* Switch kernel selection order to dotprod and i8mm

* updates for review comments

* More updates for review comments

* Reorganize and rename KleidiAI files

* Move ggml-cpu-traits.h to source file

* Update cmake for SME build and add alignment for SME

* Remove append GGML_USE_CPU_KLEIDIAI to the GGML_CDEF_PUBLIC list
This commit is contained in:
Charles Xu
2025-02-20 14:06:51 +01:00
committed by GitHub
parent 4806498bf1
commit c5d91a7400
9 changed files with 767 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
ggml/src/ggml-cpu/ggml-cpu.cpp
View File

@ -14,6 +14,10 @@
#include "ggml-cpu-hbm.h"
#endif
#ifdef GGML_USE_CPU_KLEIDIAI
#include "kleidiai/kleidiai.h"
#endif
#if defined(__APPLE__)
#include <sys/types.h>
#include <sys/sysctl.h>
@ -39,6 +43,12 @@ std::vector<ggml_backend_buffer_type_t>& ggml_backend_cpu_get_extra_buffers_type
}
#endif
#ifdef GGML_USE_CPU_KLEIDIAI
if (ggml_backend_cpu_kleidiai_buffer_type()) {
bufts.push_back(ggml_backend_cpu_kleidiai_buffer_type());
}
#endif
#ifdef GGML_USE_CPU_AARCH64
if (ggml_backend_cpu_aarch64_buffer_type()) {
bufts.push_back(ggml_backend_cpu_aarch64_buffer_type());
@ -538,6 +548,9 @@ static ggml_backend_feature * ggml_backend_cpu_get_features(ggml_backend_reg_t r
static std::string sve_cnt = std::to_string(ggml_cpu_get_sve_cnt());
features.push_back({ "SVE_CNT", sve_cnt.c_str() });
}
if (ggml_cpu_has_sme()) {
features.push_back({ "SME", "1" });
}
if (ggml_cpu_has_riscv_v()) {
features.push_back({ "RISCV_V", "1" });
}
@ -559,6 +572,9 @@ static ggml_backend_feature * ggml_backend_cpu_get_features(ggml_backend_reg_t r
#ifdef GGML_USE_OPENMP
features.push_back({ "OPENMP", "1" });
#endif
#ifdef GGML_USE_CPU_KLEIDIAI
features.push_back({ "KLEIDIAI", "1" });
#endif
#ifdef GGML_USE_CPU_AARCH64
features.push_back({ "AARCH64_REPACK", "1" });
#endif