When mixing SSE and x87 codegen, it's possible to
authorDale Johannesen <dalej@apple.com>
Mon, 24 Sep 2007 22:52:39 +0000 (22:52 +0000)
committerDale Johannesen <dalej@apple.com>
Mon, 24 Sep 2007 22:52:39 +0000 (22:52 +0000)
commit48d1e4517e6cdbbcdacae6e555a0aefa6bc3975c
treeab985187c816b4bb284c43f51e92c53a7bc4939d
parentcab5f5d269c2cf21d3ea8169b78cf62ebb87a42b
When mixing SSE and x87 codegen, it's possible to
have situations where an SSE instruction turns into
multiple blocks, with the live range of an x87
register crossing them.  To do this correctly make
sure we examine all blocks when inserting
FP_REG_KILL.  PR 1697.  (This was exposed by my
fix for PR 1681, but the same thing could happen
mixing x87 long double with SSE.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42281 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/X86/X86ISelDAGToDAG.cpp