root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 10:54:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

ggml : add max buffer sizes to opencl and metal backends (#5181)

Browse Source
This commit is contained in:
slaren 2024-01-29 09:05:13 +01:00 committed by GitHub
parent 172ac82629
commit fbe7dfa53c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 21 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
ggml-metal.m
View File

@ -2375,6 +2375,16 @@ GGML_CALL static size_t ggml_backend_metal_buffer_type_get_alignment(ggml_backen
UNUSED(buft);
}
GGML_CALL static size_t ggml_backend_metal_buffer_type_get_max_size(ggml_backend_buffer_type_t buft) {
id<MTLDevice> device = ggml_backend_metal_get_device();
size_t max_size = device.maxBufferLength;
ggml_backend_metal_free_device();
return max_size;
UNUSED(buft);
}
GGML_CALL static bool ggml_backend_metal_buffer_type_supports_backend(ggml_backend_buffer_type_t buft, ggml_backend_t backend) {
return ggml_backend_is_metal(backend) || ggml_backend_is_cpu(backend);
@ -2393,7 +2403,7 @@ GGML_CALL ggml_backend_buffer_type_t ggml_backend_metal_buffer_type(void) {
/* .get_name = */ ggml_backend_metal_buffer_type_get_name,
/* .alloc_buffer = */ ggml_backend_metal_buffer_type_alloc_buffer,
/* .get_alignment = */ ggml_backend_metal_buffer_type_get_alignment,
/* .get_max_size = */ NULL, // TODO: return device.maxBufferLength
/* .get_max_size = */ ggml_backend_metal_buffer_type_get_max_size,
/* .get_alloc_size = */ NULL, // defaults to ggml_nbytes
/* .supports_backend = */ ggml_backend_metal_buffer_type_supports_backend,
/* .is_host = */ ggml_backend_metal_buffer_type_is_host,

11
ggml-opencl.cpp
View File

@ -2125,6 +2125,15 @@ static size_t ggml_backend_opencl_buffer_type_get_alignment(ggml_backend_buffer_
GGML_UNUSED(buffer_type);
}
static size_t ggml_backend_opencl_buffer_type_get_max_size(ggml_backend_buffer_type_t buffer_type) {
static size_t max_size = -1;
if (max_size == (size_t)-1) {
ggml_cl_init();
clGetDeviceInfo(device, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(size_t), &max_size, NULL);
}
return max_size;
}
static bool ggml_backend_opencl_buffer_type_supports_backend(ggml_backend_buffer_type_t buffer_type, ggml_backend_t backend) {
//return ggml_backend_is_opencl(backend); // opencl must be used through the cpu backend
return ggml_backend_is_cpu(backend);
@ -2136,7 +2145,7 @@ static ggml_backend_buffer_type_i ggml_backend_opencl_buffer_type_interface = {
/* .get_name = */ ggml_backend_opencl_buffer_type_name,
/* .alloc_buffer = */ ggml_backend_opencl_buffer_type_alloc_buffer,
/* .get_alignment = */ ggml_backend_opencl_buffer_type_get_alignment,
/* .get_max_size = */ NULL, // TODO: return from device info
/* .get_max_size = */ ggml_backend_opencl_buffer_type_get_max_size,
/* .get_alloc_size = */ NULL,
/* .supports_backend = */ ggml_backend_opencl_buffer_type_supports_backend,
/* .is_host = */ NULL,