[Biojava-l] Benchmarks

[Matthew Pocock]

Hi all,

I have written code to count SP entries, and to convert them into Fasta
(using SeqIOListener directly and via the object model). On alphas, our
count code executes as quickly as grep (!!!). On linux, the java class
is slower by a small factor (~1.4 * grep time). The listener
implementation of the fasta maker is quicker than the one that goes via
the sequence object model (no suprises there), but I think that both are
competative with the other projects results. Anyway, you can get the
benchmarks from me and try it out yourself on some SP .dat files.

http://www.sanger.ac.uk/~mrp/benchmarks.jar

You will need to unjar this and read the README. Have fun with this.

Matthew

ps A large thanks to Thomas for writting the super-efficient IO
framework which now goes like a train on speed.