CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-09 10:17:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml-zdnn: tighten memory usage, change string allocation

Browse Source 
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-07-28 15:55:42 +08:00
parent aef93b3908
commit bee7dd3020
2 changed files with 3 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-zdnn/ggml-zdnn-impl.h
View File

@@ -59,13 +59,13 @@ typedef unsigned long long ulong64x2_t __attribute__((vector_size(16)));
struct ggml_backend_zdnn_buffer { struct ggml_backend_zdnn_buffer {
void * data; void * data;
size_t size; size_t size;
ggml_backend_zdnn_buffer * extra; // for bias etc.
zdnn_tensor_desc pre_tfm_desc; zdnn_tensor_desc pre_tfm_desc;
zdnn_tensor_desc tfm_desc; zdnn_tensor_desc tfm_desc;
zdnn_ztensor ztensor; zdnn_ztensor ztensor;
char name[GGML_MAX_NAME]; char name[GGML_MAX_NAME];
struct ggml_backend_zdnn_buffer * extra; // for bias etc.
ggml_backend_zdnn_buffer() : extra(nullptr) {} ggml_backend_zdnn_buffer() : extra(nullptr) {}
}; };

6
ggml/src/ggml-zdnn/ggml-zdnn-rewrite.cpp
View File

@@ -362,8 +362,7 @@ static enum ggml_status ggml_backend_zdnn_buffer_init_tensor(ggml_backend_buffer
tensor_buffer = &ctx->buffers[tensor_buffer_idx]; tensor_buffer = &ctx->buffers[tensor_buffer_idx];
tensor_buffer->data = tensor->data; tensor_buffer->data = tensor->data;
tensor_buffer->size = tsize; tensor_buffer->size = tsize;
strncpy(tensor_buffer->name, tensor->name, sizeof(tensor_buffer->name) - 1); snprintf(tensor_buffer->name, sizeof(tensor_buffer->name), "%s", tensor->name);
tensor_buffer->name[sizeof(tensor_buffer->name) - 1] = '\0';
ggml_zdnn_init_tensor(tensor_buffer, tensor); ggml_zdnn_init_tensor(tensor_buffer, tensor);
ctx->n_buffers++; ctx->n_buffers++;
@@ -374,8 +373,7 @@ static enum ggml_status ggml_backend_zdnn_buffer_init_tensor(ggml_backend_buffer
bias_buffer = &ctx->buffers[bias_buffer_idx]; bias_buffer = &ctx->buffers[bias_buffer_idx];
bias_buffer->data = calloc(tensor->ne[0], tensor->ne[0] * sizeof(float)); bias_buffer->data = calloc(tensor->ne[0], tensor->ne[0] * sizeof(float));
bias_buffer->size = tensor->ne[0] * sizeof(float); bias_buffer->size = tensor->ne[0] * sizeof(float);
strncpy(bias_buffer->name, "bias", sizeof(bias_buffer->name) - 1); snprintf(bias_buffer->name, sizeof(bias_buffer->name), "%s.bias", tensor->name);
bias_buffer->name[sizeof(bias_buffer->name) - 1] = '\0';
ggml_zdnn_init_bias_tensor(bias_buffer, tensor); ggml_zdnn_init_bias_tensor(bias_buffer, tensor);
ctx->n_buffers++; ctx->n_buffers++;