CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-20 12:07:33 +00:00
Code Issues Packages Projects Releases Wiki Activity

CANN: Improve ACL graph matching (#16166)

Browse Source 
* CANN: improve ACL graph matching

Record `ne` and `nb` information for src tensors and include them in the
graph matching check. This enhances the robustness of ACL graph matching
by preventing incorrect matches when src tensors share the same data
address but differ in shape or stride.

* CANN: add op_params match
This commit is contained in:
Chenguang Li
2025-10-09 15:50:25 +08:00
committed by GitHub
parent d80d6d2400
commit aa4711d369
2 changed files with 45 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
ggml/src/ggml-cann/common.h
View File

@@ -341,11 +341,18 @@ private:
#ifdef USE_ACL_GRAPH
struct ggml_graph_node_properties {
// dst tensor
void * node_address;
ggml_op node_op;
int64_t ne[GGML_MAX_DIMS];
size_t nb[GGML_MAX_DIMS];
// src tensor
void * src_address[GGML_MAX_SRC];
int64_t src_ne[GGML_MAX_SRC][GGML_MAX_DIMS];
size_t src_nb[GGML_MAX_SRC][GGML_MAX_DIMS];
// op
ggml_op node_op;
int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(int32_t)];
};