mirror of
https://github.com/ggerganov/whisper.cpp.git
synced 2025-05-22 17:10:46 +02:00
* rename ggml-cpu-aarch64.c to .cpp * reformat extra cpu backend. - clean Q4_0_N_M and IQ4_0_N_M - remove from "file" tensor type - allow only with dynamic repack - extract cpu extra bufts and convert to C++ - hbm - "aarch64" - more generic use of extra buffer - generalise extra_supports_op - new API for "cpu-accel": - amx - aarch64 * clang-format * Clean Q4_0_N_M ref Enable restrict on C++ * add op GGML_OP_MUL_MAT_ID for Q4_0_N_M with runtime repack * added/corrected control on tensor size for Q4 repacking. * Update ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * add debug logs on repacks. --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
56 lines
2.0 KiB
C++
56 lines
2.0 KiB
C++
#ifdef GGML_USE_CPU_HBM
|
|
|
|
#include "ggml-backend.h"
|
|
#include "ggml-backend-impl.h"
|
|
#include "ggml-cpu.h"
|
|
#include "ggml-impl.h"
|
|
|
|
#include "ggml-cpu-hbm.h"
|
|
|
|
// buffer type HBM
|
|
|
|
#include <hbwmalloc.h>
|
|
|
|
static const char * ggml_backend_cpu_hbm_buffer_type_get_name(ggml_backend_buffer_type_t buft) {
|
|
return "CPU_HBM";
|
|
|
|
GGML_UNUSED(buft);
|
|
}
|
|
|
|
static void ggml_backend_cpu_hbm_buffer_free_buffer(ggml_backend_buffer_t buffer) {
|
|
hbw_free(buffer->context);
|
|
}
|
|
|
|
static ggml_backend_buffer_t ggml_backend_cpu_hbm_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft,
|
|
size_t size) {
|
|
void * ptr;
|
|
int result = hbw_posix_memalign(&ptr, ggml_backend_cpu_buffer_type_get_alignment(buft), size);
|
|
if (result != 0) {
|
|
GGML_LOG_ERROR("failed to allocate HBM buffer of size %zu\n", size);
|
|
return NULL;
|
|
}
|
|
|
|
ggml_backend_buffer_t buffer = ggml_backend_cpu_buffer_from_ptr(ptr, size);
|
|
buffer->buft = buft;
|
|
buffer->iface.free_buffer = ggml_backend_cpu_hbm_buffer_free_buffer;
|
|
|
|
return buffer;
|
|
}
|
|
|
|
ggml_backend_buffer_type_t ggml_backend_cpu_hbm_buffer_type(void) {
|
|
static struct ggml_backend_buffer_type ggml_backend_cpu_buffer_type_hbm = {
|
|
/* .iface = */ {
|
|
/* .get_name = */ ggml_backend_cpu_hbm_buffer_type_get_name,
|
|
/* .alloc_buffer = */ ggml_backend_cpu_hbm_buffer_type_alloc_buffer,
|
|
/* .get_alignment = */ ggml_backend_cpu_buffer_type_get_alignment,
|
|
/* .get_max_size = */ nullptr, // defaults to SIZE_MAX
|
|
/* .get_alloc_size = */ nullptr, // defaults to ggml_nbytes
|
|
/* .is_host = */ ggml_backend_cpu_buffer_type_is_host,
|
|
},
|
|
/* .context = */ nullptr,
|
|
};
|
|
|
|
return &ggml_backend_cpu_buffer_type_hbm;
|
|
}
|
|
#endif
|