root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 02:44:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

ggml : do not print perf ops that have not been used at all

Browse Source
This commit is contained in:
Georgi Gerganov 2023-04-23 18:32:52 +03:00
parent e4422e299c
commit ec9cdb6752
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
ggml.c
View File

@ -11237,7 +11237,7 @@ void ggml_graph_print(const struct ggml_cgraph * cgraph) {
for (int i = 0; i < cgraph->n_nodes; i++) {
struct ggml_tensor * node = cgraph->nodes[i];
perf_total_per_op_us[node->op] += node->perf_time_us;
perf_total_per_op_us[node->op] += MAX(1, node->perf_time_us);
GGML_PRINT(" - %3d: [ %5" PRId64 ", %5" PRId64 ", %5" PRId64 "] %16s %s (%3d) cpu = %7.3f / %7.3f ms, wall = %7.3f / %7.3f ms\n",
i,
@ -11260,6 +11260,10 @@ void ggml_graph_print(const struct ggml_cgraph * cgraph) {
}
for (int i = 0; i < GGML_OP_COUNT; i++) {
if (perf_total_per_op_us[i] == 0) {
continue;
}
GGML_PRINT("perf_total_per_op_us[%16s] = %7.3f ms\n", GGML_OP_LABEL[i], (double) perf_total_per_op_us[i] / 1000.0);
}