commit 6b3ddbeb7da2893cae2e2fe22131d0ac8c7ebfbe
author Craig Topper <craig.topper@intel.com> Wed Apr 17 06:09:11 2019 +0000
committer Craig Topper <craig.topper@intel.com> Wed Apr 17 06:09:11 2019 +0000
tree b45a1a0cb8502c4e61ce87ebb30402076cf600c8
parent 430cf96dccd9011e71eb191ede264b71fe671ae6

[X86] In CopyToFromAsymmetricReg, use VR128 instead of FR32 instructions for GR32<->XMM register copies.

We have two versions of some instructions, VR128 versions and FR32 versions that
are marked as CodeGenOnly.

This change switches to using the VR128 versions for these copies. It's after
register allocation so the class size no longer matters. This matches how GR64
works.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@358555 91177308-0d34-0410-b5e6-96231b3b80d8