<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/64339>64339</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [regression] maybe a runtime issue related to D154953
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            new issue
      </td>
    </tr>

    <tr>
      <th>Assignees</th>
      <td>
      </td>
    </tr>

    <tr>
      <th>Reporter</th>
      <td>
          vfdff
      </td>
    </tr>
</table>

<pre>
    This is reported with @mstorsjo , so I'm checking:

This commit caused misoptimizations in the WMA decoder in ffmpeg, observed on all architectures. The misoptimization can be observed with https://martin.st/temp/wma-preproc.c, compiled with clang -target aarch64-linux-gnu -c -O3 wma-preproc.c -o libavcodec/wma.o

For a full runtime reproducible case:
```
$ git clone git://source.ffmpeg.org/ffmpeg
$ mkdir ffmpeg-build
$ cd ffmpeg-build
$ ../configure --cc=clang --samples=../fate-samples
$ make -j$(nproc)
$ make fate-rsync
$ make fate-wmapro-2ch
```
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJx0k0uPrDYQhX-N2ZRAxjy6WbCYpIWURZTNSFkbuwDP-IH86M7k10dmOt13RvdKiIeLc075k4uHoFaLOJLuN9JdCp7i5vx4XeSyFLOTH-PrpgKoAB535yNKuKm4AWmpCdH58OaAsN8hOPiDsJMBsaF4V3YlzQuhF0Lv98NFOGNUBMFTQAlGBbdHZdS_PCpnAygLcUP4-88XkCicRJ-XlsXsuOYMNwf0V5TgLHCtgXuxqYgiJo-hgtcNv3uC4BZmfCqP3rcY95D7YxNhk-E-KluFSNgU0eyETTfDy93j7p2oRE4WzuxK_68XmtsVysj9ihF4bqNvS61s-qdcbYJSQPlXA19coHSg1cyveV_iM6NyPxKanAcOS9IafLJRGYRDLJNQs0YQPOATak_v1-cna2HNZLWzmN8euwsueYHVJ8TK-ZWw6U70oTTvUvk753JOSstnTchfFKqKsEk4u6g1eYSyFII0lzuaMnCzawykuRz_LTziY-2Zy98RyjfCWsLONnMibPhWPpQ-fFjxs8LN8N27kontG5ZCjo0cmoEXONb9QIdT1zR9sY20PrcD7_tzzU99XQ-nXsoTpXTum_rMu1OhRkZZQ8-U1TVjdVMNy7Igxb6TDZ1b3pKWouFKV1pfTSZaqBASjn3bNEOh-Yw6HNPEmMUbHEXCWB4uP2ZNOac1kJZqFWJ4ukQV9TGGHlePIShnSXcBwz9mBP44E4cfeNQ8z2J0cKm7duiaInk9fj3Zq4pbmivhDGFTzrk_yt27NxT5wB9ugbDp6P6_AAAA___VU0sL">