27d15ecc2b8c8a8c9642374058ee13ba5389fd64 - llvm

commit	27d15ecc2b8c8a8c9642374058ee13ba5389fd64	[log] [tgz]
author	Craig Topper <craig.topper@intel.com>	Sun Jan 13 02:59:59 2019 +0000
committer	Craig Topper <craig.topper@intel.com>	Sun Jan 13 02:59:59 2019 +0000
tree	c5de587dc9b3927097382c452429a24245f70cdb
parent	47100f5f523424a284c2ccf3f9442179aa27d6d4 [diff]

[X86] Add more ISD nodes to handle masked versions of VCVT(T)PD2DQZ128/VCVT(T)PD2UDQZ128 which only produce 2 result elements and zeroes the upper elements.

We can't represent this properly with vselect like we normally do. We also have to update the instruction definition to use a VK2WM mask instead of VK4WM to represent this.

Fixes another case from PR34877

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@351018 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/X86/X86ISelLowering.cpp[diff]
lib/Target/X86/X86ISelLowering.h[diff]
lib/Target/X86/X86InstrAVX512.td[diff]
lib/Target/X86/X86InstrFragmentsSIMD.td[diff]
lib/Target/X86/X86IntrinsicsInfo.h[diff]

5 files changed

tree: c5de587dc9b3927097382c452429a24245f70cdb