<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/115348>115348</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [mlir] Consider moving away from linking LLVM IR for host and device code in SPIR-V CPU Runner flow
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            mlir
      </td>
    </tr>

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

    <tr>
      <th>Reporter</th>
      <td>
          andfau-amd
      </td>
    </tr>
</table>

<pre>
    See this discussion for more details: https://github.com/llvm/llvm-project/pull/114563#discussion_r1831742718

One possible outcome is we simply keep things as they are, but there may be a better option. In particular, if "device" CPU compilation/execution could be offloaded in a similar way as for GPU devices, there would be a better alignment between the SPIR-V CPU Runner and e.g. the Vulkan Runner.
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJyUUk9vnD4Q_TTmMgqyDfxgDxzyS7RVpFSNEiXXasADuDE28p_d7revzDbNoadewHjmPd68eRiCni1Rz5r_WXNfYIqL8z1aNWG6wVUVg1OX_oUI4qIDKB3GFIJ2FibnYXWeQFFEbQKrbmGJccsHJo9MHmcdlzSUo1uZPBpz-njdbN79oDEyedySMUwehaib_yomq0_-7150lWhr2YqO8XvGb6_Pb5ZgcyHowRC4FEe3EugAZ4Kg181c4J1oy2rtHAADxIUugJ6YvIMhxfztCVa8wECAMFCM5MFtUTtbwoOFDX3UYzLoM0RPwKRUdNIjMSnh7ukVRrdu2mBGMHmknzSmfIbRJaMyrZsm41CRAm0BszBt0MMZL1lRdu7L0ytcSUP-y1XU-QP_RxYaPduVbMwXZyKbO-Hl6eH55m2X8pyszX1WAZVzuZffknlH-7tUFqqv1KE6YEG9aCvBOW87Xix9Oymi6tDwbpQHPHSiU1R1XNZScikJC91LLmsheCvaSjaHcqwG3nExoBiQJmpYzWlFbcq81dL5udAhJOqFaKq6KwwOZMIeLSlXoz2TMofM93sKhjQHVnOjQwyfDFFHs8dxBzT3cOds0Ipy2E7azoDZxcm7FYy27_nm8fHtKzw8774uLsTdjau5MDpFeQl_WzYZdy6SN_0_h3YfMuyx3ec89fJXAAAA___9_BX7">