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

Update code to fix issues occuring due to non alignment of elements to be processed as multiple of 16 in MSVC

Browse source
This commit is contained in:
Srihari-mcw 2024-08-23 03:08:08 -07:00 • committed by Srihari-mcw
parent 0c81b7bcea
commit 49af3f5da7
1 changed files with 2 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml/src/ggml-aarch64.c
View file

@ -2423,7 +2423,8 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
__m256i requiredOrder = _mm256_set_epi32(3 ,2 ,1 ,0, 7 ,6, 5, 4); __m256i requiredOrder = _mm256_set_epi32(3 ,2 ,1 ,0, 7 ,6, 5, 4);
// Take group of four block_q8_0x4 structures at each pass of the loop and perform dot product operation // Take group of four block_q8_0x4 structures at each pass of the loop and perform dot product operation
for (; y < nr / 4; y += 4) { int anr = nr - nr %16; // Used to align nr with boundary of 16
for (; y < anr / 4; y += 4) {
const block_q8_0x4 * a_ptrs[4]; const block_q8_0x4 * a_ptrs[4];
a_ptrs[0] = a_ptr_start + (y * nb); a_ptrs[0] = a_ptr_start + (y * nb);