Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

hexagon: add fp16 support for binary ops: add,sub,mul,div (#20139)

Browse Source 
* hexagon: add fp16 support for binary ops: add,sub,mul,div

* hexagon: fix test-backend-ops failures for fp16 binary ops on older arches (<v79)

* hexagon: decide on n_threads (aka n_jobs) early to avoid overallocating scratchpad

* snapdragon: fix readme link

---------

Co-authored-by: Max Krasnyansky <maxk@qti.qualcomm.com>
This commit is contained in:
YardenTal44
2026-03-05 18:29:13 -08:00
committed by GitHub
co-authored by Max Krasnyansky
parent a0ed91a442
commit 2b10b62677
16 changed files with 637 additions and 378 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-hexagon/htp/set-rows-ops.c
+5 -4
View File
@@ -128,6 +128,8 @@ static void set_rows_thread_f16_f32(unsigned int nth, unsigned int ith, void *da
int op_set_rows(struct htp_ops_context * octx) {
set_rows_preamble;
const uint32_t n_threads = MIN(nr, octx->n_threads);
if (octx->src0.type != HTP_TYPE_F32) {
return HTP_STATUS_NO_SUPPORT;
}
@@ -149,15 +151,14 @@ int op_set_rows(struct htp_ops_context * octx) {
srctx.div_ne12 = init_fastdiv_values(ne12);
srctx.div_ne11 = init_fastdiv_values(ne11);
const uint32_t n_jobs = MIN(nr, octx->n_threads);
srctx.src0_nrows_per_thread = (nr + n_jobs - 1) / n_jobs;
srctx.src0_nrows_per_thread = (nr + n_threads - 1) / n_threads;
switch(octx->dst.type) {
case HTP_TYPE_F32:
worker_pool_run_func(octx->ctx->worker_pool, set_rows_thread_f32_f32, &srctx, n_jobs);
worker_pool_run_func(octx->ctx->worker_pool, set_rows_thread_f32_f32, &srctx, n_threads);
break;
case HTP_TYPE_F16:
worker_pool_run_func(octx->ctx->worker_pool, set_rows_thread_f16_f32, &srctx, n_jobs);
worker_pool_run_func(octx->ctx->worker_pool, set_rows_thread_f16_f32, &srctx, n_threads);
break;
default:
return HTP_STATUS_NO_SUPPORT;