Re: [jamboost] Optimization: testcases needed

From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2004-12-16 17:49:44

Vladimir Prus wrote:
> To be able to try different approaches quickly, I need to have testcases I can 
> play with locally, without waiting for fresh profiles from users. It's best 
> to test on Boost.Build setup used on real projects.
[snip]
> I'd also appreciate information about time "bjam -n" takes on your project and 
> hardware details (CPU, memory size, disk speed).

Hello Vladimir,

I have reworked (again) all my jamfiles (well actually most of them are 
generated by a python script) and now I think my Jamfiles are pretty 
clean and use boost-build in a correct way.

I just timed a bjam -n and on an old machine (PIII 633Mhz) this takes 
199.5 real and 178.27s user time. I also tried the same on the real 
project (where the sources files are not empty) and I was surprised that 
it took about as much time (201.89s real and 175.61 user time). This 
would indicate that parsing all files to follow all #includes would 
alsmost take no time (or does the -n option disable that?). Also the 
same test on a athlon1900 takes 114s real and 106s user-time.

I'll send the test-case tomorrow.

toon