[PATCH] D50070: [X86] Improved sched models for X86 CMPXCHG* instructions

[Andrew V. Tischenko via Phabricator via llvm-commits]

avt77 added inline comments.


================
Comment at: lib/Target/X86/X86SchedBroadwell.td:122
 
-defm : BWWriteResPair<WriteBSWAP32,[BWPort15], 1>; //
-defm : BWWriteResPair<WriteBSWAP64,[BWPort06, BWPort15], 2, [1, 1], 2>; //
+defm : BWWriteResPair<WriteCMPXCHG,[BWPort06, BWPort0156], 5, [2, 3], 5>;
+defm : BWWriteResPair<WriteBSWAP32,[BWPort15], 1>;
----------------
RKSimon wrote:
> What about the WriteCMPXCHGLd cases?
I decided (as with  BT*rr instrs) it'd be better to have separate patches for 'rr' and 'Ld' versions.The problem is:
  automaticaly generated WriteCMPXCHGLd does not fit to real numbers that's why it's better to use X86WriteRes for both 'rr' and 'Ld' versions instead of X86SchedWritePair. Should I do it in this patch?


https://reviews.llvm.org/D50070