diff options
| author | Craig Topper <craig.topper@intel.com> | 2019-03-20 21:30:20 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@intel.com> | 2019-03-20 21:30:20 +0000 |
| commit | 03675533043ddbf2ad27f260ccc410952977f574 (patch) | |
| tree | dd1b6e7e2190944a9553bef01e6a73d70e8e1a4e /llvm/lib/Object/SymbolicFile.cpp | |
| parent | bbcb95a64e186dacd66034340de81b5efc8c000e (diff) | |
| download | llvm-03675533043ddbf2ad27f260ccc410952977f574.zip llvm-03675533043ddbf2ad27f260ccc410952977f574.tar.gz llvm-03675533043ddbf2ad27f260ccc410952977f574.tar.bz2 | |
[X86] Call lowerShuffleAsBitMask for 512-bit vectors in lowerShuffleAsBlend.
This patch enables the use of lowerShuffleAsBitMask for 512-bit blends before
falling back to move immedate, GPR to k-register, and masked op.
I had to make some changes to support v8i64 when i64 is not a legal type. And to
support floating point types.
This trades a load for the move immediate and GPR move which is higher latency.
But its probably better for register pressure not having to hop through other
register classes. The load+and should play better with LICM and
rematerialization I think.
Differential Revision: https://reviews.llvm.org/D59479
llvm-svn: 356618
Diffstat (limited to 'llvm/lib/Object/SymbolicFile.cpp')
0 files changed, 0 insertions, 0 deletions