Poly/ML 4.2.0 seems to be a little slower building my stuff than version 4.1.3 generally and, if I'm not careful, a lot slower on Mac OS X.
For a typical compilation (my SLRP parser generator), I get figures like this on a Linux PC:
4.1.3: 0.84s real 0.80s user 0.02s system
4.2.0: 0.98s real 0.93s user 0.03s system
So that's about 15% slower.
On Mac OS X, the figures for the same compilation are:
4.1.3: real 0m2.323s user 0m1.439s sys 0m0.122s
4.2.0: real 0m7.84s user 0m1.96s sys 0m0.14s
So it's processing about 30% slower, but spending 70% of the real time doing nothing, so overall it's about 200% slower.
The 70% is down to an old friend, the Mac OS X crashreporter, see:
http://lists.inf.ed.ac.uk/mailman/private/polyml/2004-February/000098.html
but now with version 4.2.0, you get two or three crash-reports per run and the overhead seems to be much higher.
If you do "killall crashreporterd" before compiling, things look a lot better:
4.1.3:
real 0m1.576s user 0m1.431s sys 0m0.112s
4.2.0: real 0m2.15s user 0m1.96s sys 0m0.13s
But even with the crash-reporter overhead removed altogether, that's still more than 30% slower.
Regards,
Rob.