[PATCH] D24045: Tail Duplication: Don't denormalize loops when duplicating early.

Kyle Butt via llvm-commits llvm-commits at lists.llvm.org
Tue Aug 30 10:59:24 PDT 2016


iteratee created this revision.
iteratee added a reviewer: davidxl.
iteratee added subscribers: llvm-commits, echristo, timshen.
iteratee set the repository for this revision to rL LLVM.
Herald added a subscriber: jfb.

Duplicating loop latches, headers, or pre-headers prevents certain loop based
optimizations. Now that we have tail-duplication during placement, it makes
sense to wait to duplicate these blocks until later, to allow other
optimizations to proceed.

Repository:
  rL LLVM

https://reviews.llvm.org/D24045

Files:
  include/llvm/CodeGen/TailDuplicator.h
  lib/CodeGen/MachineBlockPlacement.cpp
  lib/CodeGen/TailDuplication.cpp
  lib/CodeGen/TailDuplicator.cpp
  test/CodeGen/WebAssembly/cfg-stackify.ll
  test/CodeGen/X86/avx512-i1test.ll
  test/CodeGen/X86/tail-dup-no-denormalize-loops.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D24045.69727.patch
Type: text/x-patch
Size: 10890 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160830/e5804900/attachment.bin>


More information about the llvm-commits mailing list