ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-30 03:11:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix last commit

Browse Source 
Did not re-check on AVX2/Zen4 after NEON related changes and,
sure enough, I broke AVX2/Zen4.
This commit is contained in:
Iwan Kawrakow
2024-10-01 14:48:44 +03:00
parent e7f5a86a41
commit 970df4b467
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/src/iqk/iqk_mul_mat.cpp
View File

@@ -7184,10 +7184,11 @@ struct HelperIQ4nl final : public BaseHelper<step> {
using Base = BaseHelper<step>; using Base = BaseHelper<step>;
#ifdef __aarch64__ #ifdef __aarch64__
using block_q8 = block_q8_0; using block_q8 = block_q8_0;
HelperIQ4nl(const char * data, int stride) : Base(data, stride), values(vld1q_s8(iq4k_values)) {}
#else #else
HelperIQ4nl(const char * data, int stride) : Base(data, stride) {}
using block_q8 = block_q8_1; using block_q8 = block_q8_1;
#endif #endif
HelperIQ4nl(const char * data, int stride) : Base(data, stride), values(vld1q_s8(iq4k_values)) {}
// Needed for v * softmax(k * q) // Needed for v * softmax(k * q)
inline void load(int l1, int i, F16::Data& v1, F16::Data& v2) const { inline void load(int l1, int i, F16::Data& v1, F16::Data& v2) const {