From: Vladimir Prus (ghost_at_[hidden])
Date: 2007-10-25 02:12:45
Motonari Ito wrote:
> The biggest problem I have in bjam is its speed.
> We have 1M lines of code managed by Boost Build system (V2). Bjam takes
> 20-30 second (when current directory is root) or 5-6 second (when
> current directory is leaf) before starting actual compilations. It's
> very slow for daily compile & edit development.
> So, I have profiled on the latest bjam (Milestone 12) with Rational
> Quantify. It looks like memory and hash related functions are bottle
> neck - is there any chance to accelerate them?
Well, the Python port of Boost.Build is progressing nicely, and
showing some nice performance number already -- in one, artificial, case,
I've got 7x speedup and 4x decrease in memory consumption as well --
and Python port is 1<->1 translation without any attempt to make proper
use of Python facilities.
More details are here:
although at this point the python port is not ready for general testing,
I expect the core files and gcc to be ready really soon.
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk