| commit | c628c1e0b40b7eca79a4f2d3a9d4befbc171eeb8 | [log] [tgz] |
|---|---|---|
| author | Michael Kuperstein <michael.m.kuperstein@intel.com> | Wed Nov 04 11:17:53 2015 +0000 |
| committer | Michael Kuperstein <michael.m.kuperstein@intel.com> | Wed Nov 04 11:17:53 2015 +0000 |
| tree | 1a262ea367b75f55c6fb25eaf14e577aab0cb00f | |
| parent | 23740f4c43d89e7e9840f3ff0a102825bfff89fa [diff] |
[X86] DAGCombine should not introduce FILD in soft-float mode The x86 "sitofp i64 to double" dag combine, in 32-bit mode, lowers sitofp directly to X86ISD::FILD (or FILD_FLAG). This should not be done in soft-float mode. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252042 91177308-0d34-0410-b5e6-96231b3b80d8