Commit 097f5b533 for llama.cpp
commit 097f5b5332f46559f1ee11d9d64e48a5065b2989
Author: Max Krasnyansky <maxk@qti.qualcomm.com>
Date: Wed Oct 7 23:48:06 2026 -0700
hex-mmadd: do not assume aligned read/write when bias-add is fused (#30133)
diff --git a/ggml/src/ggml-hexagon/htp/matmul-ops.c b/ggml/src/ggml-hexagon/htp/matmul-ops.c
index a589e0a20..1a513d160 100644
--- a/ggml/src/ggml-hexagon/htp/matmul-ops.c
+++ b/ggml/src/ggml-hexagon/htp/matmul-ops.c
@@ -425,7 +425,7 @@ static void hvx_mv_2d_repacked_##SUFFIX(unsigned int nth, unsigned int ith, void
if (copy_cnt > 0) { \
htp_trace_event_start(tr, HTP_TRACE_EVT_HVX_COMP, ct_end); \
if (src2) { \
- hvx_add_f32_uaa((uint8_t *) &dst_col[src0_start_row], \
+ hvx_add_f32_uuu((uint8_t *) &dst_col[src0_start_row], \
(const uint8_t *) tmp, \
(const uint8_t *) ((const float *) mmctx->vtcm_bias + src0_start_row), \
copy_cnt); \
@@ -1108,7 +1108,7 @@ static void hvx_mv_2d(unsigned int nth, unsigned int ith, void * data) {
if (copy_cnt > 0) {
htp_trace_event_start(tr, HTP_TRACE_EVT_HVX_COMP, src0_end_row);
if (src2) {
- hvx_add_f32_uaa((uint8_t *) &dst_col[src0_start_row],
+ hvx_add_f32_uuu((uint8_t *) &dst_col[src0_start_row],
(const uint8_t *) tmp,
(const uint8_t *) ((const float *) mmctx->vtcm_bias + src0_start_row),
copy_cnt);