Function model_map_span_vec_include_one

ds4.c:4156–4187

                isum += (int64_t)sc_val * q * (int)q8[j + 32 + l];
            }

            ql += 32;
            is += 2;
            u1 = (uint8_t)(u1 << 2);
            u2 = (uint8_t)(u2 << 2);
        }

        sumf += d * (float)isum - dmin * (float)summs;
    }

    *s = sumf;
}

static void ds4_vec_dot_q6_K_q8_K(int n, float *s, const block_q6_K *x, const block_q8_K *y) {
    const int nb = n / QK_K;
    float sumf = 0.0f;

    for (int i = 0; i < nb; i++) {
        const float d = y[i].d * f16_to_f32(x[i].d);
        const uint8_t *ql = x[i].ql;
        const uint8_t *qh = x[i].qh;
        const int8_t *scales = x[i].scales;
        const int8_t *q8 = y[i].qs;
        int64_t isum = 0;

        for (int n128 = 0; n128 < QK_K; n128 += 128) {
            for (int l = 0; l < 32; l++) {
                const int is = l / 16;
                const int q1 = ((int)(ql[l + 0]  & 0x0F) | (((qh[l] >> 0) & 3) << 4)) - 32;
                const int q2 = ((int)(ql[l + 32] & 0x0F) | (((qh[l] >> 2) & 3) << 4)) - 32;
                const int q3 = ((int)(ql[l + 0]  >> 4)    | (((qh[l] >> 4) & 3) << 4)) - 32;
                const int q4 = ((int)(ql[l + 32] >> 4)    | (((qh[l] >> 6) & 3) << 4)) - 32;

                isum += (int64_t)scales[is + 0] * q1 * (int)q8[n128 + l + 0];
                isum += (int64_t)scales[is + 2] * q2 * (int)q8[n128 + l + 32];