[patch] Simplify a vpermilps with constant mask

Jim Grosbach grosbach at apple.com
Mon Apr 21 10:47:48 PDT 2014


Why not change the clang IR codegen to just get this right in the first place?

-Jim

On Apr 21, 2014, at 7:20 AM, Rafael EspĂ­ndola <rafael.espindola at gmail.com> wrote:

> ping
> 
> On 11 April 2014 07:54, Rafael EspĂ­ndola <rafael.espindola at gmail.com> wrote:
>> With a constant mask a vpermilps is just a shufflevector. The attached
>> patch implements the simplification. This allows us to produce denser
>> code. It should also allow more folding down the line.
>> 
>> Cheers,
>> Rafael





More information about the llvm-commits mailing list