macroBenchmarks (was Re: [BUG] Wishlist for MessagTally)

Dan Ingalls Dan at SqueakLand.org
Sun Oct 7 22:05:58 UTC 2001


>    Talking of macroBenchmarks, I found it very unreliable. One day my SqM
>beat Squeak-D3D by as much as 10%, next day the situation is the other way
>around.

Hi ho, PhiHo -

The macro benchmarks do a lot of stuff.  They are sensitive to all sorts of environmental conditions such as memory available and what else may be on the screen, plus, of course, the code in the system which changes with each little update.  They trigger garbage collections, too - hey, they are benchmarks.  I never intended them as anything consistent or perfect, but they DO give relative speed numbers from one VM to another (if other things are held constant) that, in my experience, are indicative of what to expect on most large tasks (as opposed to, eg, tiny benchmarks).

This probably doesn't help, but I felt compelled to answer ;-).

Cheers
	- Dan




More information about the Squeak-dev mailing list