Speeding up ggc-simple on stage1

Jeffrey A Law law@cygnus.com
Fri Jan 14 11:38:00 GMT 2000

  In message < or7lhctraf.fsf@garnize.lsd.ic.unicamp.br >you write:
  > > Yea, you're thrashing.  The only aspect of this change that might help
  > > that is a smaller call chain within the garbage collector.
  > Which, by some of those amazing coincidences that happen just once in
  > a millenium, is exactly what the patch is trying to accomplish :-) :-)
I'd still be rather surprised.  Even at 1k levels deep in the collector you're
probably burning less than 256k of stack space unless the rs6000 port is
particularly wasteful of stack space.    Which is relatively small if
you're thrashing as badly as it seems.

  > I had expected gcc -O2 would perform tail recursion optimization in
  > all those places.  Shouldn't it?
The current tail recursion opts won't work if the function's return type
is null.  And it only works when the tail recursion is of the form
"return self (args)".  It's fairly lame.


More information about the Gcc-patches mailing list