t = vmulq_f32(vld1q_f32((float32_t*)&data_f[i]), c0_02); t = vmlaq_f32(t,vld1q_f32((float32_t*)&data_f[i+4]), c0_04); t = vmlaq_f32(t,vld1q_f32((float32_t*)&data_f[i+8]), c0_05); t = vmlaq_f32(t,vld1q_f32((float32_t*)&data_f[i+12]), ...