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

fix neon reinterpret type

Browse source
This commit is contained in:
slaren 2024-04-26 17:13:30 +02:00
parent ea591858b8
commit 41de83d5ca
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-quants.c
View file

@ -12521,7 +12521,7 @@ bool ggml_validate_row_data(enum ggml_type type, const void * data, size_t nbyte
uint32x4_t v = vld1q_u32((const uint32_t *)f + i);
uint32x4_t vexp = vandq_u32(v, vdupq_n_u32(0x7f800000));
uint32x4_t cmp = vceqq_u32(vexp, vdupq_n_u32(0x7f800000));
uint64_t mask = vget_lane_u64(vreinterpret_u64_u8(vshrn_n_u32(cmp, 8)), 0);
uint64_t mask = vget_lane_u64(vreinterpret_u64_u16(vshrn_n_u32(cmp, 8)), 0);
if (mask) {
for (size_t j = 0; j < 4; ++j) {
if (!validate_float(f[i + j], i + j)) {