[Ktechlab-devel] Studying the job...

Alan Grimes agrimes at speakeasy.net
Tue Oct 2 15:56:38 UTC 2007


Lawrence Shafer wrote:
> Alan. How's it coming? I'm willing to test even half finished stuff.

Sorry, I was to lazy to re-do that much work. =\

Putting the matrix classes in shouldn't be that hard this time around...

Unfortunately, my updates to Component have to be sync'd up with changes
made to the mainline and there are many changes which need to be made to
all of the subclasses. =|

I might do the matrix classes tonight, if I get time, but Component is a
very complex class...

> Alan Grimes wrote:
>> I've been going through a printout of Voltage Source and its related
>> classes to try to get a better understanding of how it could be
>> re-implemented in the form of a parts database entry. I keep finding
>> junk that I had removed in my own personal fork but which are still
>> present in the CVS tree... I found a tarball of what I had been passing
>> around before, it probably needs some tidying up before it gets put in
>> one of the SVN branches... If you need it, ask. -- it has problems with
>> it's engine though because it was a snapshot of the changes I was making
>> while I was rebuilding it.
>>
>> A brief profiling run on the SVN sources (using valgrind) showed that,
>> if I am interpreting this correctly, 23% of the execution time was taken
>> up in the [] methods of matrix and Vector. I'm pretty sure that my own
>> Qmatrix and Qvector classes optimized many of the 8 million calls to
>> those functions out... (and this really was a brief run...)
>>
>> Don't worry about my matrix classes messing up the engine, the results
>> should be identical until you start mucking with the LU decomposition
>> solver...

-- 
Buy Ron Paul's Money! =)
http://www.libertydollar.org/ld/ronpauldollar
Soundest investment on the planet!




More information about the Ktechlab-devel mailing list