> David, do you have comparison figures with GCC 2.8.x for the gzip > example? When the commit is done, I'll test CVS GNAT+GCC3 for correctness and performance on our number crunshing software (> 250 KSLOC) at work on x86 (Athlon, P3 and P4), SPARC, MIPS (R12k), PowerPC and report the results. -- Laurent Guerby <guerby@acm.org>