[llvm] r179316 - Optimize icmp involving addition better

[Renato Golin]

On 18 April 2013 08:49, Renato Golin <renato.golin at linaro.org> wrote:

> This is exactly what it is. I found that A8 and A9 IR for rdopt.c (where
> the segfault is) is identical, but A8 assembly works and A9 doesn't, so
> it's not just your patch. I'm reducing the problem (which is not simple)
> and will try to find the cause.
>

For no apparent reason, the test-suite started passing again...

http://lab.llvm.org:8011/builders/clang-native-arm-lnt/builds/617

Though, they're now breaking for other reasons...

http://lab.llvm.org:8011/builders/clang-native-arm-lnt/builds/623

And it seems it's your patch again... :)

I'm assuming it's the same issue, the ARM back-end being flimsy with
generic front-end changes, which is not good (nor your fault).

And here I go, investigating phantom bugs again... sigh...

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130418/931f2017/attachment.html>