<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/141925>141925</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [LLVMIR] atomicrmw unspoort half data type
        </td>
    </tr>

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

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

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

<pre>
    
ref: https://llvm.org/docs/LangRef.html#atomicrmw-instruction


does it make sense to support half atomic? 

cuda has half/half2 atomic API and has corresponding ptx
https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#atomicadd
</pre>
<img width="1" height="1" alt="" src="http://email.email.llvm.org/o/eJx8kkuL3DAMxz-NczEZHMV5HXKYZQksbKHdQ2mPGltJ3CZ28GMf3754drplL4WAArL-kv76YQhmsUQja-5Yc19giqvz4zfzM9FE5kei4uL028jE2dPM6jNfYzwCq88MJgbTtj3vJ-cXBpN2KjCYHtEuTzSf1rhvDGqMbjfK7y-lsSH6pKJxlolz_qBizd3Djgux5p5B_1l6MXFNl5NyO4MpBfIlxohq3cnG3AhDoOuPmAUJRFGS6rtSDliVPRCWs-p7wlaCHhSD4b2pdhS4iXzH38QD2UA8Oh7ScTgf-YrbzN9HZvXE30tU0shXDNcsgykHuL3i568PHK2-5pXznsLhrDZ24Ud8ZeL8eafs0ck-G23wtlgWv4VSlYd3i8d9N3Ypl2Q0MZiM1fT62U7UutBjrYd6wILGqpN9UzV1VRXr2F2Gtp3z4l3dkZSqUyi0rsUghVR9X5gRBDSigaESsmqqU9tdtJznugU5y7ZFJgXtaLbT3-MWJoREYyWrAZpiwwtt4QoMgKUXfs0ygMyPH3NReUlLYFJsJsTwTyaauF1Je3z8_uXhiTX3_AMPnmw43McNNEbk8e2gIvlt_A8YWfwWsnu_SMXsWR4ps3Gb-XmEPwEAAP__hQjwHw">