Speed improvements and slowups 3.0.x to latest 3.1.x

[David Bateman]

To see if there were any speed regressions I just tried running the 
sciview benchmark on a 3.0.x version of Octave and comparing it against 
a build from the repository from a hour a ago. I got the results 
attached. The good news is that

* 3.1 is faster for transposes
* 3.1 is even faster than the above would explain for operations 
involving transposes like a' * a, though it appears that a \ b' doesn't 
get an advantage from this
* The Escoufier test is much faster, I presume due to the recent 
indexing fixes from Jaroslav..

The bad news is that the sorting appears to be slower.. Now there have 
been several changes to the sorting that were made, that major one from 
me that made the sort a method of the array classes. However, I seem to 
remember at the time I did this that I didn't notice significant speed 
changes.. I know Jaroslav replaced some malloc/free coe in the sort code 
with new/delete and a couple of other things... Could this be the cause? 
Any ideas how to get the speed of 3.0.x back for this test?

D.

-- 
David Bateman                                dbateman at dbateman.org
35 rue Gambetta                              +33 1 46 04 02 18 (Home)
92100 Boulogne-Billancourt FRANCE            +33 6 72 01 06 33 (Mob)

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: sciview.3_0_3
Url: https://www-old.cae.wisc.edu/pipermail/octave-maintainers/attachments/20090206/3d6214a3/attachment-0002.ksh 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: sciview.3_1_51+
Url: https://www-old.cae.wisc.edu/pipermail/octave-maintainers/attachments/20090206/3d6214a3/attachment-0003.ksh