[patch][pr22947] Use sext in fast isel

[Eric Christopher]

Hi Rafael,

The changes to the mips fast isel test are worrying. From looking at the
testcase it's materializing a constant 1 and you have it now materializing
a constant -1 (extended) into the register.

-eric

On Mon, Apr 6, 2015 at 12:00 PM Rafael Espíndola <rafael.espindola at gmail.com>
wrote:

> ping
>
> On 2 April 2015 at 13:28, Rafael Espíndola <rafael.espindola at gmail.com>
> wrote:
> > ping
> >
> > On 27 March 2015 at 14:00, Rafael Espíndola <rafael.espindola at gmail.com>
> wrote:
> >> ping
> >>
> >> On 18 March 2015 at 11:45, Rafael Espíndola <rafael.espindola at gmail.com>
> wrote:
> >>> Fast isel currently has a bug in that it zero extends immediates to 64
> >>> bits. This normally goes unnoticed because the value is truncated to
> >>> 32 bits for output.
> >>>
> >>> Two cases were it is noticed:
> >>>
> >>> * We fail to use smaller encodings.
> >>> * If the original constant was smaller than i32.
> >>>
> >>> For example, the only possible values for i1 are 0 and -1, yet we were
> >>> expanding -1 to i32 1.
> >>>
> >>> Cheers,
> >>> Rafael
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150406/424f2dda/attachment.html>