[PATCH] Allow loop vectorization with llvm.lifetime calls

Jessome, Marc marc.jessome at intel.com
Wed Jul 31 07:58:21 PDT 2013


Hello,

This patch will allow vectorization of loops which contain calls to llvm.lifetime start/end. In order to preserve lifetime information across the loop, that call to llvm.lifetime.* is emit into the vectorized loop body.

In this patch, the special-casing of these intrinsic calls was done since there is no ready way to:  1. differentiate non-vectorized calls during loop body construction, and 2. differentiate Intrinsic calls vs. intrinsic instructions while costing. Perhaps someone with a bit more experience in the area could point me in a better direction here.

Marc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130731/9ae90c70/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vectorize_lifetime.1.patch
Type: application/octet-stream
Size: 4844 bytes
Desc: vectorize_lifetime.1.patch
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130731/9ae90c70/attachment.obj>


More information about the llvm-commits mailing list