vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml : move AMX to the CPU backend (#10570)

Browse source 
* ggml : move AMX to the CPU backend

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Diego Devesa 2024-11-29 21:54:58 +01:00 • committed by GitHub
parent b782e5c7d4
commit 7cc2d2c889
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
64 changed files with 514 additions and 801 deletions
Download patch file Download diff file Expand all files Collapse all files

15
ggml/src/ggml-cpu/ggml-cpu-impl.h
View file

@ -15,6 +15,18 @@
extern "C" {
#endif
struct ggml_compute_params {
// ith = thread index, nth = number of threads
int ith, nth;
// work buffer for all threads
size_t wsize;
void * wdata;
struct ggml_threadpool * threadpool;
};
#if defined(_MSC_VER)
#define m512bh(p) p
@ -366,6 +378,9 @@ static __m256 __lasx_xvreplfr2vr_s(float val) {
}
#endif
// TODO: move to ggml-threading
void ggml_barrier(struct ggml_threadpool * tp);
#ifdef __cplusplus
}
#endif