[PATCH] D52886: [X86] Move ReadAfterLd functionality into X86FoldableSchedWrite (PR36957)

[Clement Courbet via Phabricator via llvm-commits]

courbet added inline comments.


================
Comment at: lib/Target/X86/X86SchedSandyBridge.td:78
+def : ReadAdvance<ReadAfterVecXLd, 6>;
+def : ReadAdvance<ReadAfterVecYLd, 7>;
+
----------------
Looks good, except that store-to-load forwarding seems to incur an extra cycle in the `ymm` case.


================
Comment at: lib/Target/X86/X86SchedSkylakeServer.td:78
 
+// Vector loads are 5/6/7 cycles, so ReadAfterVec*Ld registers needn't be available
+// until 5/6/7 cycles after the memory operand.
----------------
I'm seeing 4.5(??)/5/6-15/15/16 here.


Repository:
  rL LLVM

https://reviews.llvm.org/D52886