[llvm] r211888 - [x86] Begin a significant overhaul of how vector lowering is done in the

Chandler Carruth chandlerc at gmail.com
Fri Jun 27 10:20:54 PDT 2014


On Fri, Jun 27, 2014 at 7:12 PM, Nadav Rotem <nrotem at apple.com> wrote:

> The vectorizers don’t generate interesting shuffle patterns because it is
> difficult to predict the cost of shuffles. We use the ShuffleKind enum in
> TTI to query the backends on the cost of specific instructions such as
> reverse, broadcast and and alternate. For loop vectorization I don’t think
> that there are other interesting shuffle patterns. Can you think of other
> patterns that can help loop vectorization? For SLP-vectorization, shuffling
> loads could help, but the problem is cost estimation and not inefficient
> lowering.


I don't know what to tell you, as I'm not familiar with the shuffles in the
vectorizer, however I observe that we vectorize loops and the results end
up in the x86 shuffle lowering, with rather interesting patterns. Some of
these patterns are the reason I started looking at all of this.

:: shrug ::
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140627/ebf673c0/attachment.html>


More information about the llvm-commits mailing list