From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-04-22 09:19:58
----- Original Message -----
From: "Hamish Mackenzie" <hamish_at_[hidden]>
If you are wondering why these figures are better than the
> ones for size then you are not alone. It seems that the compiler
> prefers the input data I used here which is a repeating sequence of
> different classes (for size I tested a sequence containing only one
> type). I am going to look into this some more.
have you read my instantiation speed paper? We found some awfully
counter-intuitive results for GCC during that project also.
Unfortunately it can be difficult to even determine what to measure.
BTW, the paper contains some attempts to account for irrelevant effects;
you might want to read about it because it may lay some groundwork for
getting meaningful numbers.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk