Need for Speed (postscript)

Stephen Barncard stephenREVOLUTION at barncard.com
Sun Jul 15 16:50:16 EDT 2007


Has someone benchmarked this?

I could easily stand corrected on this (not tested by me), but I'd 
look for a way to touch the variable outPutData "by reference" (in 
other words touching as little as possible).

  "Put into", I think, has more overhead. In a situation with so much 
data, every little bit helps.

I know that in HC, when you put somevariable into tOutVariable, that 
the entire tOutVariable gets recreated, not just appended.

Ken Ray has a demonstration of the the use of "by reference" but does 
not allude to any speed improvement.

http://www.sonsothunder.com/devres/revolution/tips/scrp003.htm


Also again I could be corrected, but perhaps moving the data from 
field "massList" into a variable might speed things up a bit. I'm 
wondering that ANY referring to any field from inside such a long 
loop is a possible time consumer.

another thing -- the last line in the loop

put "====================" & return after outputData

perhaps that could be a constant, defined at the top of the script


constant kBarDisplay = "===================="

then use this in the loop
put kBarDisplay & return after outputData

this one might have negligible difference, but I haven't tested this.


>Thanks to everyone who responded (Mark Schonewille, Mark Smith, 
>Viktoras Didziulis, Robert Brenstein, apologies if I forgot anyone) 
>- a whole range of solutions, including arrays, databases and 'just' 
>variables. Anyway, I coded up the brute force method without any 
>finess... here's the business bit of the script with some frippery...
>
>
>
>
>         put "NEW SEARCH, MASS = " & pepMass & " at " & ppm & " ppm 
>error" & return after outputData
>
>



>
>
>I ran a brutal test, of 45,000 lines in massList and 16,000 lines in seqDB
>
>My crude attempt seems to be capable, even running within the Rev 
>IDE, of completing the 720million comparisons in about 30minutes 
>(OK, admittedly CoreDuo 2.66GHz, 2GB RAM). That's 24million a 
>minute! (I deliberately put some searches that would match at the 
>end of seqDB, to be sure I searched through most of the file each 
>time). I am pretty happy with this, and I'd be looking for at least 
>a 10-fold gain in speed to code up a harder solution.
>
>
>
>Do you experts thing a 10-fold gain is feasible? 100-fold?
>
>Rob
>
>________________________________
>
>Prof R J Beynon[h]
>Proteomics and Functional Genomics Group
>Faculty of Veterinary Science
>University of Liverpool
>Crown Street, Liverpool L69 7ZJ
>
>________________________________
>
>Phone: +44 151 794 4312
>
>Fax: +44 151 794 4243
>
>Email: r.beynon at liv.ac.uk
>
>http://www.liv.ac.uk/pfg
>
>________________________________
>
>
>
>This email was sent on Sun, 15 Jul, 2007 at 8:58 PM.
>
>
>
>_______________________________________________
>use-revolution mailing list
>use-revolution at lists.runrev.com
>Please visit this url to subscribe, unsubscribe and manage your 
>subscription preferences:
>http://lists.runrev.com/mailman/listinfo/use-revolution


-- 


stephen barncard
s a n  f r a n c i s c o
- - -  - - - - - - - - -






More information about the Use-livecode mailing list