[llvm-commits] [llvm] r49003 - in /llvm/trunk/test/CodeGen: PowerPC/fold-li.ll PowerPC/reg-coalesce-simple.ll PowerPC/rotl-2.ll X86/2006-03-02-InstrSchedBug.ll X86/2006-05-02-InstrSched1.ll X86/2006-05-02-InstrSched2.ll X86/2006-05-11-InstrSched.ll X86/constant-pool-remat-0.ll X86/iabs.ll X86/or-branch.ll X86/select.ll X86/setuge.ll X86/x86-64-mem.ll X86/zero-remat.ll

[Duncan Sands]

> The labels in question all seem to be valid when EH is being generated.

I meant: in a function with no invokes, does turning on eh (now the default)
generate labels that wouldn't have been generated with eh turned off?  The
problem with labels is that they get in the way of some codegen optimizations.
It would be bad if by turning on eh unconditionally, code that doesn't require
eh is being slowed down.

Ciao,

Duncan.