llvm dev - Dec 2009 - [LLVMdev] 2.5 Regalloc Assert

I am hitting this assert in LLVM 2.5:

/// spillPhysRegAroundRegDefsUses - Spill the specified physical register
/// around all defs and uses of the specified interval.
void LiveIntervals::spillPhysRegAroundRegDefsUses(const LiveInterval &li,
                                            unsigned PhysReg, VirtRegMap
&vrm)
{
  unsigned SpillReg = getRepresentativeReg(PhysReg);

  for (const unsigned *AS = tri_->getAliasSet(PhysReg); *AS; ++AS)
    // If there are registers which alias PhysReg, but which are not a
    // sub-register of the chosen representative super register. Assert
    // since we can't handle it yet.
    assert(*AS == SpillReg || !allocatableRegs_[*AS] ||
           tri_->isSuperRegister(*AS, SpillReg));

The test also asserts on trunk, though with an error about running out
of registers.  I am working to file a bug.

I am also going to try 2.6.

Any hints on what might be going on?

                               -Dave

On Dec 7, 2009, at 3:46 PM, David Greene wrote:
> I am hitting this assert in LLVM 2.5:
> 
> /// spillPhysRegAroundRegDefsUses - Spill the specified physical register
> /// around all defs and uses of the specified interval.
> void LiveIntervals::spillPhysRegAroundRegDefsUses(const LiveInterval
&li,
>                                            unsigned PhysReg, VirtRegMap
&vrm)
> {
>  unsigned SpillReg = getRepresentativeReg(PhysReg);
> 
>  for (const unsigned *AS = tri_->getAliasSet(PhysReg); *AS; ++AS)
>    // If there are registers which alias PhysReg, but which are not a
>    // sub-register of the chosen representative super register. Assert
>    // since we can't handle it yet.
>    assert(*AS == SpillReg || !allocatableRegs_[*AS] ||
>           tri_->isSuperRegister(*AS, SpillReg));
> 
> The test also asserts on trunk, though with an error about running out
> of registers.  I am working to file a bug.
The allocator cannot find a virtual register interval to spill. So it's
attempting to slice up a physical register interval to free up a register around
the uses and defs of interval li.

Looks like it's unable to find a suitable register to do so.

Evan
> 
> I am also going to try 2.6.
> 
> Any hints on what might be going on?
> 
>                               -Dave
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev