CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-08 10:07:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml-zdnn: attempt to fix sigsegv

Browse Source 
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-07-28 23:37:50 +08:00
parent e0549c2925
commit fc9260deab
1 changed files with 7 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
ggml/src/ggml-zdnn/ggml-zdnn-rewrite.cpp
View File

@@ -201,12 +201,6 @@ static enum ggml_status ggml_zdnn_graph_compute(ggml_backend_t backend, ggml_cgr
}
static bool ggml_zdnn_supports_op(const ggml_backend_zdnn_device_context * ctx_dev, const ggml_tensor * op) {
const ggml_tensor * src0 = op->src[0];
const ggml_tensor * src1 = op->src[1];
const ggml_tensor * dst = op;
GGML_TENSOR_BINARY_OP_LOCALS
switch (op->op) {
case GGML_OP_NONE:
case GGML_OP_RESHAPE:
@@ -217,6 +211,13 @@ static bool ggml_zdnn_supports_op(const ggml_backend_zdnn_device_context * ctx_d
case GGML_OP_MUL_MAT:
{
const ggml_tensor * src0 = op->src[0];
const ggml_tensor * src1 = op->src[1];
const int64_t ne10 = src1->ne[0];
const int64_t ne0 = op->ne[0];
const int64_t ne1 = op->ne[1];
const int64_t max_batch = zdnn_get_nnpa_max_dim_idx_size();
return ggml_is_contiguous(src0) &&