projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 662024d) | patch
[X86] When expanding LCMPXCHG16B_NO_RBX in EmitInstrWithCustomInserter, directly...
authorCraig Topper <craig.topper@intel.com>
Fri, 9 Oct 2020 18:48:10 +0000 (11:48 -0700)
committerCraig Topper <craig.topper@intel.com>
Fri, 9 Oct 2020 18:55:24 +0000 (11:55 -0700)
commitf34bb06935aa3bab353d70d515b767fdd2f5625c
tree832b247172354f39b2e35ea782c4f246d37f9f0dtree | snapshot
parent662024df331bd1f1a206678435e51232683e3cf6commit | diff
[X86] When expanding LCMPXCHG16B_NO_RBX in EmitInstrWithCustomInserter, directly copy address operands instead of going through X86AddressMode.

I suspect getAddressFromInstr and addFullAddress are not handling
all addresses cases properly based on a report from MaskRay.

So just copy the operands directly. This should be more efficient
anyway.
llvm/lib/Target/X86/X86ISelLowering.cpp diff | blob | history
Domain: System / Toolchain;