register renaming causes many ARM testsuite failures

[Jeffrey A Law]

  In message <Pine.LNX.4.30.0101021721550.24769-100000@host140.cambridge.redhat
.com>you write:
  > On Sat, 30 Dec 2000, Philip Blundell wrote:
  > 
  > > Many test cases are failing to compile at -O3 using the trunk sources on 
  > ARM
  > > targets, with errors like this:
  > >
  > > /tmp/cc4c42x1.s: Assembler messages:
  > > /tmp/cc4c42x1.s:21: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:23: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:24: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:25: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:28: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:31: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:57: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:59: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:60: Error: Bad range in register list
  > > /tmp/cc4c42x1.s:61: Error: Bad range in register list
  > >
  > > The problem is that an insn like this matches the special `load multiple'
  > > pattern which generates an LDM instruction:
  > 
  > Apparently the constraints on the ldmsi patterns aren't strict enough.
Actually, I would think it would be the operand predicates or the insn's
condition, not the constraints.
jeff