On Thu, Aug 14, 2025 at 5:42 PM 钟居哲 <juzhe.zh...@rivai.ai> wrote:
>
> OK. How about this following situation:
>
> for (i ... i < 64.. i++) {
> a[i] = a[i] + b[i]
> }
>
> When TARGET_MAX_LMUL is 1 or 2. Can auto-vectorize use LMUL 8 in VLS mode to
> vectorize the case above ?
OK, let me add one more testcase and I also found a bug at
vectorize_related_mode for VLS mode, it will result in V16SI for
V16QI, but it should be V4SI.
Will send updated patch for adding testcase and also fixing for
vectorize_related_mode
