From: Kawrakow Date: Mon, 24 Jul 2023 09:55:02 +0000 (+0300) Subject: Fix scalar version of Q5_K when QK_K = 64 (#2362) X-Git-Tag: gguf-v0.4.0~410 X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=42f70cb2f6a8089e0a0560a459e4ba317bac4d49;p=pkg%2Fggml%2Fsources%2Fllama.cpp Fix scalar version of Q5_K when QK_K = 64 (#2362) Co-authored-by: Iwan Kawrakow --- diff --git a/k_quants.c b/k_quants.c index c576fd7a..e790abf8 100644 --- a/k_quants.c +++ b/k_quants.c @@ -3297,8 +3297,7 @@ void ggml_vec_dot_q5_K_q8_K(const int n, float * restrict s, const void * restri #else - - uint8_t aux8[QK_K]; + int8_t aux8[QK_K]; int16_t aux16[16]; float sums [8]; memset(sums, 0, 8*sizeof(float)); @@ -3308,7 +3307,7 @@ void ggml_vec_dot_q5_K_q8_K(const int n, float * restrict s, const void * restri const uint8_t * restrict q4 = x[i].qs; const uint8_t * restrict hm = x[i].qh; const int8_t * restrict q8 = y[i].qs; - uint8_t * restrict a = aux8; + int8_t * restrict a = aux8; for (int l = 0; l < 32; ++l) { a[l+ 0] = q4[l] & 0xF; a[l+32] = q4[l] >> 4;