[PATCH] D12006: [X86][SSE] Use the general SMAX/SMIN/UMAX/UMIN pattern matching and remove the X86 implementation

[Simon Pilgrim via llvm-commits]

RKSimon added a comment.

In http://reviews.llvm.org/D12006#223805, @qcolombet wrote:

>




> Just asking, since I haven’t followed the previous related patches: Aren’t we regressing AVX512 by removing matchIntegerMINMAX?

>  More generally, does the generic code produce as many MIN/MAX patterns as this target specific one?


Before this patch I'll put in some more thorough min max tests to ensure that we don't (there might be a few AVX512 cases that aren't been declared legal yet). I was intending to do this when I added constant folding support anyway.


Repository:
  rL LLVM

http://reviews.llvm.org/D12006