<table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Issue</th>
<td>
<a href=https://github.com/llvm/llvm-project/issues/153375>153375</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>
[VectorCombine] Investigate using TCK_CodeSize for minsize builds
</td>
</tr>
<tr>
<th>Labels</th>
<td>
llvm::vectorcombine
</td>
</tr>
<tr>
<th>Assignees</th>
<td>
</td>
</tr>
<tr>
<th>Reporter</th>
<td>
RKSimon
</td>
</tr>
</table>
<pre>
VectorCombine uses a global CostKind to drive the cost:benefit of the possible folds, although this is currently always TCK_RecipThroughput, but it does make it easy to experiment withe the other cost kinds.
For MinSize/-Os builds we should investigate using TCK_CodeSize instead.
NOTE: Not all targets are great at realistically modelling CostKinds other than TCK_RecipThroughput so this might unearth some weaknesses in some cost models.
</pre>
<img width="1" height="1" alt="" src="http://email.email.llvm.org/o/eJx8Uk1v2zgQ_TXUZRCDpiI7OujgddbAItgESIK9LihxLE5DkQZnZNf99QXlFEWBogdB5HA-3nvzLDONEbFTzV-qeazsLD7l7vXpjaYUqz65a_cfDpLyPk09RYSZkcHCGFJvA-wTyxNFB5LAZTojiEcYEouqdz1GPJJAOi7RU2KmPiAcU3CszB5sEJ_m0YN4YiCGYc4Zo4Qr2HCxV4b3_dP_rzjQ6d3nknmapRT2swAJuIQMk_3AckHL1wIDv54w04RR4EJlbvmSeMwLLvig6Hil9E7p3SFl-JfiG31DZQ53Lwz9TMExXBDYpzk4oHhGFhqtFO4UxwXTPjksVUCRBa377Pf88v63qnfwnARsCCA2jygMNiOMGa2AFchoA7HQYEO4wpQchlD6_tCSP9GKt_F3AgCnm2ATjV5gjmizeOA0IVzQfkTksiOKt9BCepnCq8p1tWvr1lbYrbdNo7eb1tSV745bXW_bZqPNYCxq3az7473WbYtH7WpnKuqMNo1-WNe6NUbXq3bd4IPVm7VrUW_6rbrXOFkKqxDO0yrlsSLmGbt1U9fbpgq2x8CLzYwpKareqXp3Xrw13LyljCkezF15v-vnkdW9LlLxz6ZCEha3_uJK1TzCP39e1DFlmChyOd-WXM05dF7kxAWLOShzGEn83K-GNClzWFDefnennL7gIMocFlaszOGT2Lkz3wMAAP__jP0fbg">