performance hit with functions of vectors

[augustm]

Dear David,

the information you provided was most useful.

1) Firstly why is the vector sparse? I am calculating it as

octave> zz=diag(u); 

where u is from a LU factorization. The thing that was
shocking me was the discovery that performing the full factorization

octave> [l u p q]= lu(M);

was much slower than working with the diagonal elements of the result.
I need the the log-function for an application in quantum
mechanics.


Note, there are no elements of value zero in the vector. They are all positive
and "non extreme"  0.25<zz<146

however converting to full does work to speed things up.

octave:1> tic; j=log(zz); toc
Elapsed time is 0.58444 seconds.

octave:2> tic; j=log(full(zz)); toc
Elapsed time is 0.0019059 seconds.


There is thus a "delicate" interaction between sparse behaviour and
function application.

But Thank You again for putting me on the right track!

TM,
Paris 05.



      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page. 
http://www.yahoo.com/r/hs