[patch] Use add32ri8 and friends on fast iself

[Eric Christopher]

On Fri, Mar 13, 2015 at 3:02 PM Chris Lattner <clattner at apple.com> wrote:

> On Mar 13, 2015, at 2:48 PM, Eric Christopher <echristo at gmail.com> wrote:
>
> This seems reasonable to me. I'm not sure of the rationale behind making
> the selector smaller - to speed it up a bit by giving it fewer options?
>
> Chris? Here's some archaeology for you :)
>
>
> IIRC the issue was that fastisel didn’t know to call the predicate to
> check that they were valid.  That facility got added later.  Using these
> instructions is definitely goodness.
>
>
Sweet. Thanks for the brain picking :)

Rafael - have at.

-eric


> -Chris
>
>
> -eric
>
> On Fri, Mar 13, 2015 at 2:29 PM Rafael Espíndola <
> rafael.espindola at gmail.com> wrote:
>
>> This fixes pr22854.
>>
>> The core issue on the bug is that there are multiple instructions that
>> print the same in assembly. In fact, there doesn't seem to be any
>> syntax
>> for specifying that a constant that fits in 8 bits should use a 32 bit
>> immediate.
>>
>> The attached patch changes fast isel to consider i16immSExt8,
>> i32immSExt8, and i64immSExt8. It is interesting to note that these
>> predicates were explicitly disabled in r129692. It is not clear
>> exactly what the issue was.
>>
>> Other options would be:
>>
>> * Make MC at -O0 (-mc-relax-all) relax constants. That would use the
>> larger encoding even when parsing assembly.
>> * Invent a syntax for requiring a 32 bit immediate.
>>
>> Cheers,
>> Rafael
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150313/0abd5d1c/attachment.html>