Re: m68k/regclass/regmove fun

[Jeffrey A Law <law at upchuck dot cygnus dot com>]

  In message <36FFFEA2.33C67526@moene.indiv.nluug.nl>you write:
  > Jeffrey A Law wrote:
  > 
  > [ ... DAXPY code elided ... ]
  > 
  > > Toon -- if you could give this patch a benchmark spin it would be greatly
  > > appreciated:
  > 
  > Ah, I often wondered why it did this ...
Because you're a sadistic bastard :-) :-)

  > Unfortunately, benchmarking it would take a while, because building two
  > versions of a recent snapshot on my NeXTStation (assuming nothing goes
  > wrong) takes about 20 hours - so I tend to procrastinate this into the
  > weekend :-)
  > 
  > But the reasoning seems sound, so why not install it ?
Well, I was hoping to get some data, particularly from a benchmark other
than daxpy that what I was doing isn't absurd :-)  I'm pretty sure the daxpy
loop will improve :-)


How long does it take to build just a stage1/cc1 and stage1/f771 without
debug or optimization?  That's typically what I do to test this kind of little
stuff quickly on my slower machines.

jeff

ps.  Count me in for the Fortran BOF.