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

fix: use vm_allocate to allocate CPU backend buffer on macOS

Browse source
This commit is contained in:
Gilad S 2024-10-14 02:30:25 +03:00
parent d4c19c0f5c
commit 06b119a177
1 changed files with 11 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

11
ggml/src/ggml-backend.cpp
View file

@ -24,6 +24,8 @@
#ifdef __APPLE__ #ifdef __APPLE__
#include <sys/types.h> #include <sys/types.h>
#include <sys/sysctl.h> #include <sys/sysctl.h>
#include <mach/mach.h>
#include <TargetConditionals.h>
#endif #endif
@ -770,12 +772,21 @@ static const char * ggml_backend_cpu_buffer_type_get_name(ggml_backend_buffer_ty
} }
static ggml_backend_buffer_t ggml_backend_cpu_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) { static ggml_backend_buffer_t ggml_backend_cpu_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
#ifdef TARGET_OS_OSX
void * data = NULL;
kern_return_t alloc_status = vm_allocate((vm_map_t) mach_task_self(), (vm_address_t *) &data, size, VM_FLAGS_ANYWHERE);
if (alloc_status != KERN_SUCCESS) {
GGML_LOG_ERROR("%s: failed to allocate buffer using vm_allocate with size %zu\n", __func__, size);
return NULL;
}
#else
size += TENSOR_ALIGNMENT; // malloc may return an address that is not aligned size += TENSOR_ALIGNMENT; // malloc may return an address that is not aligned
void * data = malloc(size); // TODO: use GGML_ALIGNED_MALLOC (move to ggml-impl.h) void * data = malloc(size); // TODO: use GGML_ALIGNED_MALLOC (move to ggml-impl.h)
if (data == NULL) { if (data == NULL) {
GGML_LOG_ERROR("%s: failed to allocate buffer of size %zu\n", __func__, size); GGML_LOG_ERROR("%s: failed to allocate buffer of size %zu\n", __func__, size);
return NULL; return NULL;
} }
#endif
return ggml_backend_buffer_init(buft, ggml_backend_cpu_buffer_i, data, size); return ggml_backend_buffer_init(buft, ggml_backend_cpu_buffer_i, data, size);
} }