vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

adding the current_chunk

Browse source
This commit is contained in:
Kunnis 2024-05-09 22:12:20 -05:00
parent bb1b1d0071
commit daa87b1813
1 changed files with 4 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml.c
View file

@ -1578,6 +1578,8 @@ struct ggml_compute_state_shared {
ggml_abort_callback abort_callback; // abort ggml_graph_compute when true ggml_abort_callback abort_callback; // abort ggml_graph_compute when true
void* abort_callback_data; void* abort_callback_data;
atomic_int current_chunk; // currently processing chunk during Mat_Mul, shared between all the threads.
}; };
struct ggml_compute_state { struct ggml_compute_state {
@ -12003,6 +12005,8 @@ UseGgmlGemm1:;
if (ith != 0) { if (ith != 0) {
return; return;
} }
//Every thread starts at ith, so the first unprocessed chunk is nth. This save a bit of coordination right at the start.
atomic_store(&state->shared->current_chunk, nth);
if (src1->type != vec_dot_type) { if (src1->type != vec_dot_type) {
char * wdata = params->wdata; char * wdata = params->wdata;
const size_t row_size = ggml_row_size(vec_dot_type, ne10); const size_t row_size = ggml_row_size(vec_dot_type, ne10);