diff options
| author | Craig Topper <craig.topper@intel.com> | 2019-06-21 17:24:21 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@intel.com> | 2019-06-21 17:24:21 +0000 |
| commit | 6af1be96641f34e10bf3b4866f72571b63fab27c (patch) | |
| tree | 5316df5a0d97530c01855f9f7d587b6109a68927 /llvm/lib/Target/WebAssembly/WebAssemblyFixFunctionBitcasts.cpp | |
| parent | 4c9def4a51ac10d9a249f31ea712c32474e89914 (diff) | |
| download | llvm-6af1be96641f34e10bf3b4866f72571b63fab27c.zip llvm-6af1be96641f34e10bf3b4866f72571b63fab27c.tar.gz llvm-6af1be96641f34e10bf3b4866f72571b63fab27c.tar.bz2 | |
[X86] Use vmovq for v4i64/v4f64/v8i64/v8f64 vzmovl.
We already use vmovq for v2i64/v2f64 vzmovl. But we were using a
blendpd+xorpd for v4i64/v4f64/v8i64/v8f64 under opt speed. Or
movsd+xorpd under optsize.
I think the blend with 0 or movss/d is only needed for
vXi32 where we don't have an instruction that can move 32
bits from one xmm to another while zeroing upper bits.
movq is no worse than blendpd on any known CPUs.
llvm-svn: 364079
Diffstat (limited to 'llvm/lib/Target/WebAssembly/WebAssemblyFixFunctionBitcasts.cpp')
0 files changed, 0 insertions, 0 deletions