gcc/resource.c mark_target_live_regs Question

[Jeffrey A Law]

 In message < 20010215094346K.mitchell@codesourcery.com >you write:
  > Why is it still safe, then, to go back all the way to the previous
  > BARRIER and then use the global_live_at_start information for the
  > basic block that starts at this point?
I don't know the precise details (I'd have to sit down a re-read jump
and reorg).  But the basic idea is that the state at the barrier is
known to be stable for some reason.  We can go forward from there.

jeff