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: test ztensor finding in init_tensor

Browse Source 
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-07-24 00:40:22 +08:00
parent af9f4f0039
commit 9e84742e72
1 changed files with 30 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
ggml/src/ggml-zdnn/ggml-zdnn-rewrite.cpp
View File

@@ -230,11 +230,36 @@ static void * ggml_backend_zdnn_buffer_get_base(ggml_backend_buffer_t buffer) {
} }
static enum ggml_status ggml_backend_zdnn_buffer_init_tensor(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor) { static enum ggml_status ggml_backend_zdnn_buffer_init_tensor(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor) {
// TODO: init tensor here struct ggml_backend_zdnn_buffer_context * ctx = (struct ggml_backend_zdnn_buffer_context *)buffer->context;
return GGML_STATUS_SUCCESS;
GGML_UNUSED(buffer); // Find which buffer entry this tensor belongs to
GGML_UNUSED(tensor); const int64_t tsize = ggml_nbytes(tensor);
size_t offset = 0;
int buffer_idx = -1;
for (int i = 0; i < ctx->n_buffers; ++i) {
const int64_t ioffs = (int64_t)tensor->data - (int64_t)ctx->buffers[i].data;
if (ioffs >= 0 && ioffs + tsize <= (int64_t)ctx->buffers[i].size) {
buffer_idx = i;
offset = (size_t)ioffs;
break;
}
}
if (buffer_idx >= 0) {
// Copy tensor name to buffer for logging
strncpy(ctx->buffers[buffer_idx].name, tensor->name, sizeof(ctx->buffers[buffer_idx].name) - 1);
ctx->buffers[buffer_idx].name[sizeof(ctx->buffers[buffer_idx].name) - 1] = '\0';
GGML_LOG_INFO("%s: initialized tensor '%s' in buffer %d, size = %8.2f MiB, offset = %zu\n",
__func__, ctx->buffers[buffer_idx].name, buffer_idx,
(float)tsize / (1024.0f * 1024.0f), offset);
} else {
GGML_LOG_WARN("%s: could not find buffer for tensor '%s'\n", __func__, tensor->name);
return GGML_STATUS_FAILED;
}
return GGML_STATUS_SUCCESS;
} }
static void ggml_backend_zdnn_buffer_memset_tensor(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor, uint8_t value, size_t offset, size_t size) { static void ggml_backend_zdnn_buffer_memset_tensor(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor, uint8_t value, size_t offset, size_t size) {
@@ -400,7 +425,7 @@ static struct ggml_backend_i ggml_backend_zdnn_i = {
}; };
static ggml_guid_t ggml_backend_zdnn_guid(void) { static ggml_guid_t ggml_backend_zdnn_guid(void) {
static char * guid_str = "IBM-ZDNN_ACCELER"; static const char * guid_str = "IBM-ZDNN_ACCELER";
return reinterpret_cast<ggml_guid_t>((void *)guid_str); return reinterpret_cast<ggml_guid_t>((void *)guid_str);
} }