ggml : fix bug in Q4_1 x Q8_1 I8MM kernel

ggml-ci

Author: ggerganov

ggml/src/ggml-cpu/ggml-cpu-quants.c

@@ -1792,7 +1792,8 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
                 const int8x16_t y1_l = vld1q_s8(b_y1->qs);
                 const int8x16_t y1_h = vld1q_s8(b_y1->qs + 16);
 
-                float32_t _scale[4] = { GGML_FP16_TO_FP32(b_x0->d)*GGML_FP16_TO_FP32(b_y0->d),
+                float32_t _scale[4] = {
+                    GGML_FP16_TO_FP32(b_x0->d)*GGML_FP16_TO_FP32(b_y0->d),
                     GGML_FP16_TO_FP32(b_x0->d)*GGML_FP16_TO_FP32(b_y1->d),
                     GGML_FP16_TO_FP32(b_x1->d)*GGML_FP16_TO_FP32(b_y0->d),
                     GGML_FP16_TO_FP32(b_x1->d)*GGML_FP16_TO_FP32(b_y1->d)};
@@ -2357,10 +2358,12 @@ void ggml_vec_dot_q4_1_q8_1(int n, float * restrict s, size_t bs, const void * r
                 const block_q8_1 * restrict b_y0 = &vy0[i];
                 const block_q8_1 * restrict b_y1 = &vy1[i];
 
-                float32_t summs_t[4] = {GGML_FP16_TO_FP32(b_x0->m) * GGML_FP16_TO_FP32(b_y0->s),
+                float32_t summs_t[4] = {
+                    GGML_FP16_TO_FP32(b_x0->m) * GGML_FP16_TO_FP32(b_y0->s),
                     GGML_FP16_TO_FP32(b_x1->m) * GGML_FP16_TO_FP32(b_y0->s),
                     GGML_FP16_TO_FP32(b_x0->m) * GGML_FP16_TO_FP32(b_y1->s),
                     GGML_FP16_TO_FP32(b_x1->m) * GGML_FP16_TO_FP32(b_y1->s)};
+
                 summs0 = vaddq_f32(summs0, vld1q_f32(summs_t));
 
                 const uint8x16_t m4b = vdupq_n_u8(0x0F);
@@ -2381,10 +2384,11 @@ void ggml_vec_dot_q4_1_q8_1(int n, float * restrict s, size_t bs, const void * r
                 const int8x16_t y1_h = vld1q_s8(b_y1->qs + 16);
 
                 // mmla into int32x4_t
-                float32_t _scale[4] = {GGML_FP16_TO_FP32(b_x0->d)*b_y0->d,
-                    GGML_FP16_TO_FP32(b_x0->d)*b_y1->d,
-                    GGML_FP16_TO_FP32(b_x1->d)*b_y0->d,
-                    GGML_FP16_TO_FP32(b_x1->d)*b_y1->d};
+                float32_t _scale[4] = {
+                    GGML_FP16_TO_FP32(b_x0->d)*GGML_FP16_TO_FP32(b_y0->d),
+                    GGML_FP16_TO_FP32(b_x0->d)*GGML_FP16_TO_FP32(b_y1->d),
+                    GGML_FP16_TO_FP32(b_x1->d)*GGML_FP16_TO_FP32(b_y0->d),
+                    GGML_FP16_TO_FP32(b_x1->d)*GGML_FP16_TO_FP32(b_y1->d)};
                 float32x4_t scale = vld1q_f32(_scale);
 
                 int8x16_t l0 = vreinterpretq_s8_s64(vzip1q_s64(vreinterpretq_s64_s8(x0_l), vreinterpretq_s64_s8(x1_l)));