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

Interleave threads

Browse source
This commit is contained in:
Will Beddow 2023-04-07 17:00:37 -04:00
parent 1b6fd5470b
commit 5e5a653555
1 changed files with 7 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

29
ggml.c
View file

@ -4938,12 +4938,6 @@ static void ggml_compute_forward_dup_f16(
const int thread_num = params->ith;
const int total_threads = params->nth;
const int64_t regions = ne03 * ne02 * ne01 * ne00;
const int64_t regions_per_thread = (regions + total_threads - 1) / total_threads;
const int64_t thread_start_region = regions_per_thread * thread_num;
const int64_t thread_stop_region = MIN(thread_start_region + regions_per_thread, regions);
int region_index = 0;
@ -4952,9 +4946,9 @@ static void ggml_compute_forward_dup_f16(
for (int64_t i02 = 0; i02 < ne02; i02++) {
for (int64_t i01 = 0; i01 < ne01; i01++) {
for (int64_t i00 = 0; i00 < ne00; i00++) {
if (region_index > thread_stop_region) break;
if (region_index++ >= thread_start_region) {
// Interleave execution so that in a 4 thread run thread 0 copies regions 0,4,8, ...
if ((region_index++ % total_threads) == thread_num) {
const char * src0_ptr = ((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
char * dst_ptr = ((char *) dst->data + i10*nb0 + i11*nb1 + i12*nb2 + i13*nb3);
@ -4984,9 +4978,8 @@ static void ggml_compute_forward_dup_f16(
for (int64_t i01 = 0; i01 < ne01; i01++) {
for (int64_t i00 = 0; i00 < ne00; i00++) {
if (region_index > thread_stop_region) break;
if (region_index++ >= thread_start_region) {
// Interleave execution so that in a 4 thread run thread 0 copies regions 0,4,8, ...
if ((region_index++ % total_threads) == thread_num) {
const char * src0_ptr = ((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
char * dst_ptr = ((char *) dst->data + i10*nb0 + i11*nb1 + i12*nb2 + i13*nb3);
@ -5053,12 +5046,6 @@ static void ggml_compute_forward_dup_f32(
const int thread_num = params->ith;
const int total_threads = params->nth;
const int64_t regions = ne03 * ne02 * ne01 * ne00;
const int64_t regions_per_thread = (regions + total_threads - 1) / total_threads;
const int64_t thread_start_region = regions_per_thread * thread_num;
const int64_t thread_stop_region = MIN(thread_start_region + regions_per_thread, regions);
int region_index = 0;
@ -5067,9 +5054,9 @@ static void ggml_compute_forward_dup_f32(
for (int64_t i02 = 0; i02 < ne02; i02++) {
for (int64_t i01 = 0; i01 < ne01; i01++) {
for (int64_t i00 = 0; i00 < ne00; i00++) {
if (region_index > thread_stop_region) break;
if (region_index++ >= thread_start_region) {
// Interleave execution so that in a 4 thread run thread 0 copies regions 0,4,8, ...
if ((region_index++ % total_threads) == thread_num) {
const char * src0_ptr = ((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
char * dst_ptr = ((char *) dst->data + i10*nb0 + i11*nb1 + i12*nb2 + i13*nb3);
@ -5098,9 +5085,7 @@ static void ggml_compute_forward_dup_f32(
for (int64_t i02 = 0; i02 < ne02; i02++) {
for (int64_t i01 = 0; i01 < ne01; i01++) {
for (int64_t i00 = 0; i00 < ne00; i00++) {
if (region_index > thread_stop_region) break;
if (region_index++ >= thread_start_region) {
if ((region_index++ % total_threads) == thread_num) {
const char * src0_ptr = ((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
char * dst_ptr = ((char *) dst->data + i10*nb0 + i11*nb1 + i12*nb2 + i13*nb3);