Here is a function in C, you have to convert it to WASM SIMD 128. This function is prone to produce inaccuracy result, it is very sensitive. Only optimize part that you are absolutely sure. Think carefully about it. Hint, focus more on the loop with aux16, aux32 ``` void ggml_vec_dot_q6_K_q8_K(int n, float * restrict s, size_t bs, const void * restrict vx, size_t bx, const void * restrict vy, size_t by, int nrc) { assert(n % QK_K == 0); assert(nrc == 1); UNUSED(nrc); UNUSED(bx); UNUSED(by); UNUSED(bs); const block_q6_K * restrict x = vx; const block_q8_K * restrict y = vy; const int nb = n / QK_K; int8_t aux8[QK_K]; int16_t aux16[8]; float sums [8]; int32_t aux32[8]; memset(sums, 0, 8*sizeof(float)); float sumf = 0; for (int i = 0; i < nb; ++i) { const uint8_t * restrict q4 = x[i].ql; const uint8_t * restrict qh = x[i].qh; const int8_t * restrict q8 = y[i].qs; memset(aux32, 0, 8*sizeof(int32_t)); int8_t * restrict a = aux8; for (int j = 0; j < QK_K; j += 128) { for (int l = 0; l < 32; ++l) { a[l + 0] = (int8_t)((q4[l + 0] & 0xF) | (((qh[l] >> 0) & 3) << 4)) - 32; a[l + 32] = (int8_t)((q4[l + 32] & 0xF) | (((qh[l] >> 2) & 3) << 4)) - 32; a[l + 64] = (int8_t)((q4[l + 0] >> 4) | (((qh[l] >> 4) & 3) << 4)) - 32; a[l + 96] = (int8_t)((q4[l + 32] >> 4) | (((qh[l] >> 6) & 3) << 4)) - 32; } a += 128; q4 += 64; qh += 32; } a = aux8; int is = 0; for (int j = 0; j < QK_K/16; ++j) { int scale = x[i].scales[is++]; for (int l = 0; l < 8; ++l) aux16[l] = q8[l] * a[l]; for (int l = 0; l < 8; ++l) aux32[l] += scale * aux16[l]; q8 += 8; a += 8; for (int l = 0; l < 8; ++l) aux16[l] = q8[l] * a[l]; for (int l = 0; l < 8; ++l) aux32[l] += scale * aux16[l]; q8 += 8; a += 8; } const float d = GGML_FP16_TO_FP32(x[i].d) * y[i].d; for (int l = 0; l < 8; ++l) sums[l] += d * aux32[l]; } for (int l = 0; l < 8; ++l) sumf += sums[l]; *s = sumf; } ```