Nonlocal gotos: wrong frame pointer handling

[Jim Wilson]

Andreas Krebbel wrote:
> A blockage is defined in the s390 backend as an UNSPEC_VOLATILE. Therefore the scheduler can't 
> exchange the insns. But this solution is apparently platform dependent and therefore is not the right way. 
> My question is how to generate a blockage which is platform independent or whether there is a better way 
> to solve this.

The MEM clobbers emitted by emit_stack_restore should do the trick.  We 
already have the clobbers emitted after the instruction that sets the 
stack pointer.  We also need them before the instruction to solve this 
problem.

The scheduler could perhaps be smart enough to know that explicit stores 
to the stack pointer/frame pointer will clobber memory, but this is so 
rare that it may not be worthwhile to handle this in the scheduler.

You could also fix the problem by defining your own nonlocal_goto 
pattern, but it would be nicer to fix this for all targets.  So adding 
the MEM clobbers seems like a better solution.
-- 
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com