PS: Building Octave on cygwin with gcc3.3.3

[Paul Thomas]

I put a lot of effort into trying to understand why octave was suffering so 
badly from the problem with setjmp.  Using timing comparisons for octave, 
with and without SjLj, and the measured times for new/delete, I concluded 
that octave was spending about half of its time(for interpreter based 
operation), with good gcc, allocating and deallocating memory.

I overloaded new/delete at runtime to do a count and therefore to estimate 
the timing implications.  The figure that I came to that way was about 10%. 
Profiling, inaccurate though it is, shows that as much as 70% of octave's 
time is spent dealing with SjLj.

Thus, I came to the conclusion that some other part of gcc than new/delete 
is primarily responsible. I am beginning, from other non-octave indications, 
to SUSPECT virtual functions and their Vtable accesses as the culprit.  I 
intend to use a model of the octave classes to do comparative timing tests.

Paul Thomas 




-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------