tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-17 13:40:55 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml : improve graph build time via hash table lookup (#2329)

Browse Source 
* improve graph build time

* ggml_tensor : use 1 bit per flag

* use a hash table instead
This commit is contained in:
slaren
2023-07-25 14:32:20 +02:00
committed by GitHub
parent 82552b7f54
commit da1889834a
3 changed files with 42 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
ggml.h
View File

@@ -442,7 +442,7 @@ extern "C" {
void * extra; // extra things e.g. for ggml-cuda.cu
char padding[8];
char padding[4];
};
static const size_t GGML_TENSOR_SIZE = sizeof(struct ggml_tensor);
@@ -463,6 +463,11 @@ extern "C" {
void * abort_callback_data;
};
// next prime after GGML_MAX_NODES
// #define GGML_GRAPH_HASHTABLE_SIZE 4099
// next prime after GGML_MAX_NODES * 2 (nodes + leafs)
#define GGML_GRAPH_HASHTABLE_SIZE 8273
// computation graph
struct ggml_cgraph {
int n_nodes;
@@ -472,6 +477,8 @@ extern "C" {
struct ggml_tensor * grads[GGML_MAX_NODES];
struct ggml_tensor * leafs[GGML_MAX_NODES];
void * visited_hash_table[GGML_GRAPH_HASHTABLE_SIZE];
// performance
int perf_runs;
int64_t perf_cycles;