Re: [boost] Scons users?

From: David Abrahams (dave_at_[hidden])
Date: 2005-09-20 10:46:56

Phillip Seaver <phil_at_[hidden]> writes:

> Martin Bonner wrote:
>
>> Well, without the numbers, I agree. But the numbers do rather back him up
>> (except that BoostBuild does seem to be marginally quicker than Scons on
>> three out of four tests, whereas he describes it as "Slower even than
>> Scons!").
>>
>> Are there any known optimization oportunities in BoostBuild?
>>
>> Are the results an artifact of his test set up?
>
> It looks to me like he's still using m10 (unless it was a typo or 
> failure to update), and Volodya said this about the numbers the first time:
>
>>However, things are not as bad as it seem, because your test is artificial.
>>It has a *lot* of similarly named targets, and that hits some non-scalable 
>>part in Boost.Build. Fixing that dropped the time quite significally, though
>>that's not in CVS yet so I can't give exact data.

Oh, _that's_ both ridiculous and inexcusable. It's hard to take his
numbers seriously, except that they are going to have an effect on
public opinion.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com