[llvm] r278658 - [LSR] Don't try and create post-inc expressions on non-rotated loops

[Reid Kleckner via llvm-commits]

On Mon, Aug 15, 2016 at 11:48 AM, James Molloy <James.Molloy at arm.com> wrote:

> Surprised I didn’t get failmail about this - is it because WebAssembly is
> an experimental target?
>

Yep. Last time I raised the idea of promoting it, people felt that the spec
was still too fluid to bless the current backend.

Given the number of tests I've had to fix in it, I'm starting to feel like
we should promote it anyway. If people build code with it and rely on a
spec that's still fluid, they get what they pay for. The "experimental"
designation has more to do with LLVM internal continuous integration anyway.


> Test update is probably the right thing to do - when people write loops
> manually in IR they tend to write them head-tested, which would trigger
> this heuristic change. It’s simply changing from using a post-inc value to
> a pre-inc value which decreases register pressure and is an all-round good
> thing, but does cause some test churn.
>

OK, updated in r278713.


> Again, I’m worried that I didn’t notice this… :/
>

Don't worry about it, it's not currently enabled. :)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160815/1bd9d01f/attachment.html>