C Optimization, Pentium 4, 4 May 2004, tree-ssa/3.5/3.4/icc

[Scott Robert Ladd]

Diego Novillo wrote:
> Ah, OK.  Thanks.  You are positive that tree-ssa gets the right result
> on alma, eh?  I wonder if we are hitting a micro-optimization of some
> kind.  Can I get the source code for these tests somewhere?

When I compile with -DVERIFY, the alma produces identical output when 
generated by tree-ssa and 3.4. Of course, the I/O time dominates the 
performance profile, so tree-ssa is only slightly faster with VERIFY 
defined. It could be, I suppose, the tree-ssa and Intel optimize away 
most of the code when nothing is output. I'll look into it...

-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing