Re: a profiler for cwm

Le jeu 29/07/2004 à 10:28, Dominique Hazaël-Massieux a écrit :
> Impressive indeed, it took 2 minutes on my machine, that is almost twice
> than before! It looks like the resulting file has some bugs; I'll check
> on my side, since this likely due to the changes in the rules I made,
> but I'll let you know if not. (the errors are buggy empty blank nodes,
> FWIW)

So, reverting to the rules as they are in the current production version
of TR automation not only fixed the buggy empty blank nodes problem, but
cut the processing time down to 1min25! 

(I think the rules I'm using shouldn't produce this problem of empty
blank nodes, but that's a different bug anyway; I'll try to make a test
case to illustrate it)

Vivien, it's probably worth up upgrading both your set up and the Web
instructions to use this version (tagged rel-0-8-0-perf) instead of
'fast_for_dom'.

Thanks again, Joseph!

Dom
-- 
Dominique Hazaël-Massieux - http://www.w3.org/People/Dom/
W3C/ERCIM
mailto:dom@w3.org