[PATCH] [x86] set default reciprocal (division and square root) codegen to match GCC

[Sanjay Patel]

Hi hfinkel, ahatanak, nadav, andreadb,

D8982 ( checked in at http://reviews.llvm.org/rL239001 ) added command-line options to specify reciprocal estimate codegen.

This patch changes the default settings for x86 targets to allow recip codegen (except for scalar division because that breaks too much code) when using -ffast-math or its equivalent. This will match GCC behavior for this kind of codegen.

http://reviews.llvm.org/D10396

Files:
  lib/Target/X86/X86TargetMachine.cpp
  test/CodeGen/X86/recip-fastmath.ll
  test/CodeGen/X86/sqrt-fastmath.ll

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D10396.27519.patch
Type: text/x-patch
Size: 7335 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150611/d61a81cf/attachment.bin>