[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [help-GIFT] Re: Testing GIFT-modifications

From: risc
Subject: Re: [help-GIFT] Re: Testing GIFT-modifications
Date: Wed, 16 Aug 2006 12:30:13 -0500
User-agent: Mutt/1.4.1i

On Wed, Aug 16, 2006 at 06:14:27PM +0200, Jonas Lindqvist wrote:
> Hi again!
> (This email is really directed to address@hidden, but following  
> Julia's example, I'll cc the help-gift list, in case someone else is  
> interested)
> >I'm just modifying the feature extraction program, so I have a small
> >shell script that runs the extractor from upstream, and then diffs
> >that against the result of 'my' extractor, 40 or so times.
> Yeah, that would be a good way to do it I suppose. I guess one could  
> write a quick hack that starts a server and sends various queries to  
> it, to make sure the response is the same over and over again... I  
> don't know what queries are really relevant though...
> >I've religiously avoided C++, as I've seen too many college grads
> >who think the whole world fits their object model, and who write
> >*HORRIBLE* performing applications. to say nothing of how they look.
> You can easily tell the difference between an experienced C++  
> programmer and one that just believes he's advanced by looking at how  
> many "free functions" and template functions he's using (especially  
> in anonymous namespaces, and such). Over-designing and thinking  
> "everything" needs to be an object is common, and it's really bad for  
> performance. Objects are great for some things though. But you can  
> write object oriented code in C if you want to. But C++ adds some  
> really neat stuff to the language besides classes, so I like using C+ 
> + for this kind of programs.

I've written a lot of object oriented C. even looked at objC for a bit. :)

> >I justify it to myself by knowing I've got leaps and bounds of
> >improvements to add to the C feature extractor.
> How do you find the bottlenecks? Some profiler?

oprofile works great. ;)

> >After todays commit, CVS should run at 1.42 on my machine,
> >VS 0.1.14 running at 1.9, and my 'private' branch running at .85!
> Sounds great... Is the private branch about to be committed any time  
> soon? I'm going to index some 90000 images any week now, but would  
> rather wait for a faster extractor if it's coming... :-)

the 'private' tree is a copy i hacked to little-bits, tried everything,
and stiched back together. its completely unsuitable for comittal, as there
are dozens of things i tried, that didnt work, but are still in it.

at the moment, my work on gift consists of isolating the things
that DO work from my private branch, writing them over again, and
commiting to mainline. tedious work, but i've found tricks while
doing that that actually sped up my private branch.

If you're needing to do a mass import like that, i can send
you/the_list a copy of my private branch, for temporary purposes.
when i'm done pulling it apart, mainline should be even faster!

> >I'm tired of spending hours waiting on my images to import, how
> >about you? ;)
> Hours? Days... :-)

Days? the actual application i have for gift could take years. :)

> /Jonas

Julia Longtin <address@hidden>

reply via email to

[Prev in Thread] Current Thread [Next in Thread]