Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 13:58:46 +01:00
Code Issues Packages Projects Releases Wiki Activity

ggml : always allocate buffers with size multiple of GGML_MEM_ALIGN

Browse Source
This commit is contained in:
Georgi Gerganov 2023-04-14 13:31:15 +03:00
parent 723dac55fa
commit c14e0d2f23
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
ggml.c
View File

@ -3054,9 +3054,11 @@ struct ggml_context * ggml_init(struct ggml_init_params params) {
return NULL;
}
const size_t mem_size = (params.mem_size + GGML_MEM_ALIGN - 1) & ~(GGML_MEM_ALIGN - 1);
*ctx = (struct ggml_context) {
/*.mem_size =*/ params.mem_size,
/*.mem_buffer =*/ params.mem_buffer ? params.mem_buffer : GGML_ALIGNED_MALLOC(params.mem_size),
/*.mem_size =*/ mem_size,
/*.mem_buffer =*/ params.mem_buffer ? params.mem_buffer : GGML_ALIGNED_MALLOC(mem_size),
/*.mem_buffer_owned =*/ params.mem_buffer ? false : true,
/*.no_alloc =*/ params.no_alloc,
/*.n_objects =*/ 0,
@ -3066,7 +3068,7 @@ struct ggml_context * ggml_init(struct ggml_init_params params) {
/*.scratch_save =*/ { 0, 0, NULL, },
};
GGML_ASSERT(ctx->mem_buffer != NULL); // check for allocation failure
GGML_ASSERT(ctx->mem_buffer != NULL);
ggml_assert_aligned(ctx->mem_buffer);