[Biococoa-dev] Translation

John Timmer jtimmer at bellatlantic.net
Fri Sep 3 13:59:12 EDT 2004


I've checked in some updated code, still using codons but having moved a
bunch of things into arrays, rather than using BCSequenceDNA objects.

The time is down by a bit, but not a whole lot (maybe about 10-15% - it's
just under .4 seconds now).  ObjC_msgSend is down to about 23%, and testing
whether a base in the codon represents one in the sequence is up to about
17.5%.  Since that's the most important step, I view this as a bit of
progress.

I'm not sure if there's any other obvious ways for me to squeeze performance
out of this - please look over the code and offer suggestions.

JT

 
> 
> I had a brief window where bioinformatics.org appeared, and I added a bunch
> of stuff then, so the code may be visible.  It requires a different .plist
> format which I didn't manage to upload, though, so it is currently
> non-functional.
> 
> Shark turned out to be incredibly easy to use.  I just looped through the
> translation 100 times while sampling, and out popped a list of what's being
> done sorted by percent of CPU use.  About 28% of the time is being used by
> ObjC_msgSend, another 17% or so with array index related stuff.
> 
> This isn't really surprising, given that we've injected another object layer
> (the codons) into the process.  I'm thinking I might be able to remove the
> BCSequenceDNA from the codon and just store all three bases as variables,
> and pull the array out of the sequence being translated and deal with the
> base array directly - both of these should cut down on the messaging
> substantially.
> 
> Hopefully, a positive report later this evening -
> 
> John
> 
> _______________________________________________
> This mind intentionally left blank
> 
> 

_______________________________________________
This mind intentionally left blank





More information about the Biococoa-dev mailing list