Re: About the last patch for private variables

[Boris Kolpackov]

Hi Paul,

Paul D. Smith <address@hidden> writes:

> Also everyone please test this and make sure I didn't break anything.
> Also I think that the new algorithm is not significantly slower than the
> previous one but it's probably worth testing on any massive makefiles
> you happen to have lying around (esp. on makefiles with lots of
> variables assignments).

I just tested current CVS with my build system (which is quite
complex) and noticed about ~30% slowdown compared to 3.81. Both
were built with full optimization. Here are the results of the 
three make-cvs runs:

real 1.05
user 0.70
sys 0.35

real 1.04
user 0.73
sys 0.29

real 1.06
user 0.75
sys 0.30

And here are the three make-3.81 runs:

real 0.71
user 0.63
sys 0.04

real 0.68
user 0.63
sys 0.05

real 0.71
user 0.65
sys 0.04

Notice the difference in the system times. I can run it under oprofile
and show the timings or I can provide instructions on how to reproduce
it (the software package that is built is open-source).

Boris 

-- 
Boris Kolpackov, Code Synthesis Tools   http://codesynthesis.com/~boris/blog
Open source XML data binding for C++:   http://codesynthesis.com/products/xsd
Mobile/embedded validating XML parsing: http://codesynthesis.com/products/xsde