[PATCH] D86883: [X86] Add support for using fast short rep mov for memcpy lowering.

Craig Topper via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Sep 2 09:57:47 PDT 2020


craig.topper added inline comments.


================
Comment at: llvm/lib/Target/X86/X86SelectionDAGInfo.cpp:198
   SDValue InFlag;
-  Chain = DAG.getCopyToReg(Chain, dl, CX, Size, InFlag);
+  Chain = DAG.getCopyToReg(
+      Chain, dl, CX,
----------------
yamauchi wrote:
> craig.topper wrote:
> > Is this needed because we're no longer calling this with just constants we fixed the size of?
> Do you mean, "is the getCopyToReg call needed"? The size needs to be loaded into the RCX/ECX register for the rep movs instruction and the callers aren't doing that. So, it seems yes. Did you mean in a different way?
I was asking why the getZExtOrTrunc was added. do we create memcpy's where the size type isn't the same size as the pointer?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D86883/new/

https://reviews.llvm.org/D86883



More information about the llvm-commits mailing list