CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-28 08:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

metal : remove deprecated ggml_backend_metal_buffer_from_ptr

Browse Source
This commit is contained in:
Georgi Gerganov
2025-09-09 09:28:41 +03:00
parent bdff7729b1
commit d91ba85d04
2 changed files with 0 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/include/ggml-metal.h
View File

@@ -43,10 +43,6 @@ GGML_BACKEND_API ggml_backend_t ggml_backend_metal_init(void);
GGML_BACKEND_API bool ggml_backend_is_metal(ggml_backend_t backend);
GGML_DEPRECATED(
GGML_BACKEND_API ggml_backend_buffer_t ggml_backend_metal_buffer_from_ptr(void * data, size_t size, size_t max_size),
"obsoleted by the new device interface - https://github.com/ggml-org/llama.cpp/pull/9713");
GGML_BACKEND_API void ggml_backend_metal_set_abort_callback(ggml_backend_t backend, ggml_abort_callback abort_callback, void * user_data);
GGML_BACKEND_API ggml_backend_buffer_type_t ggml_backend_metal_buffer_type(void);

89
ggml/src/ggml-metal/ggml-metal.m
View File

@@ -6143,95 +6143,6 @@ static ggml_backend_buffer_type_t ggml_backend_metal_buffer_from_ptr_type(void)
return &ggml_backend_buffer_from_ptr_type_metal;
}
// TODO: obsoleted by ggml_backend_metal_device_buffer_from_ptr
ggml_backend_buffer_t ggml_backend_metal_buffer_from_ptr(void * data, size_t size, size_t max_size) {
struct ggml_backend_metal_buffer_context * ctx = calloc(1, sizeof(struct ggml_backend_metal_buffer_context));
ctx->all_data = data;
ctx->all_size = size;
ctx->owned = false;
ctx->n_buffers = 0;
const size_t size_page = sysconf(_SC_PAGESIZE);
// page-align the data ptr
{
const uintptr_t offs = (uintptr_t) data % size_page;
data = (void *) ((char *) data - offs);
size += offs;
}
size_t size_aligned = size;
if ((size_aligned % size_page) != 0) {
size_aligned += (size_page - (size_aligned % size_page));
}
struct ggml_backend_metal_device_context * ctx_dev = &g_ggml_ctx_dev_main;
GGML_ASSERT(ctx_dev->mtl_device != nil);
id<MTLDevice> device = ctx_dev->mtl_device;
// the buffer fits into the max buffer size allowed by the device
if (size_aligned <= device.maxBufferLength) {
ctx->buffers[ctx->n_buffers].data = data;
ctx->buffers[ctx->n_buffers].size = size;
ctx->buffers[ctx->n_buffers].metal = nil;
if (size_aligned > 0) {
ctx->buffers[ctx->n_buffers].metal = [device newBufferWithBytesNoCopy:data length:size_aligned options:MTLResourceStorageModeShared deallocator:nil];
if (ctx->buffers[ctx->n_buffers].metal == nil) {
GGML_LOG_ERROR("%s: error: failed to allocate buffer, size = %8.2f MiB\n", __func__, size_aligned / 1024.0 / 1024.0);
return false;
}
}
ggml_backend_metal_log_allocated_size(device, size_aligned);
++ctx->n_buffers;
} else {
// this overlap between the views will guarantee that the tensor with the maximum size will fully fit into
// one of the views
const size_t size_ovlp = ((max_size + size_page - 1) / size_page + 1) * size_page; // round-up 2 pages just in case
const size_t size_step = device.maxBufferLength - size_ovlp;
const size_t size_view = device.maxBufferLength;
for (size_t i = 0; i < size; i += size_step) {
const size_t size_step_aligned = (i + size_view <= size) ? size_view : (size_aligned - i);
ctx->buffers[ctx->n_buffers].data = (void *) ((uint8_t *) data + i);
ctx->buffers[ctx->n_buffers].size = size_step_aligned;
ctx->buffers[ctx->n_buffers].metal = nil;
if (size_step_aligned > 0) {
ctx->buffers[ctx->n_buffers].metal = [device newBufferWithBytesNoCopy:(void *) ((uint8_t *) data + i) length:size_step_aligned options:MTLResourceStorageModeShared deallocator:nil];
if (ctx->buffers[ctx->n_buffers].metal == nil) {
GGML_LOG_ERROR("%s: error: failed to allocate buffer, size = %8.2f MiB\n", __func__, size_step_aligned / 1024.0 / 1024.0);
return false;
}
}
ggml_backend_metal_log_allocated_size(device, size_step_aligned);
if (i + size_step < size) {
GGML_LOG_INFO("\n");
}
++ctx->n_buffers;
}
}
if (!ggml_backend_metal_buffer_rset_init(ctx, ctx_dev, device)) {
GGML_LOG_ERROR("%s: error: failed to initialize residency set\n", __func__);
free(ctx);
return NULL;
}
return ggml_backend_buffer_init(ggml_backend_metal_buffer_from_ptr_type(), ggml_backend_metal_buffer_i, ctx, size);
}
// backend
static const char * ggml_backend_metal_name(ggml_backend_t backend) {