From maintainers-request at octave dot org Tue Jan 10 07:30:07 2006
Subject: Re: Successful compilation with MinGW
From: Paul Kienzle <pkienzle at users dot sourceforge dot net>
To: octave maintainers mailing list <maintainers at octave dot org>
Date: Tue, 10 Jan 2006 08:28:48 -0500


On Jan 10, 2006, at 5:57 AM, David Bateman wrote:

> John Swensen wrote:
>
>> John W. Eaton wrote:
>> > On  9-Jan-2006, David Bateman wrote:
>> >
>> > | So if you want a version of octave under windows that won't be
>> > | embrassing slow on for-loops in particular you need mingw...
>> >
>> > Or you need to use dwarf exception handling with Cygwin, or you need
>> > someone to find out what is causing the performance problem with  
>> sjlj
>> > exception handling and fix that problem.
>> >
>> > jwe
>> >
>> >
>> >  Maybe I am being a little naive about how deep the Octave exception
>> handling is used, but from a pretty minimal glance it appears that the
>> exception handling is fairly confined as to where exceptions are  
>> thrown
>> and caught.  Would it be an option to somehow implement a "custom"
>> exception handler and avoid C++ exceptions altogether?
>>
>> John
>>
> The macro OCTAVE_QUIT occurs all over the Octave code to handle the  
> user hitting ctrl-c deep in a calculation. So you have the posibility  
> of having an exception at every occurence of this macro deep within  
> the inner loops of the calculations. This means that all the memory  
> that it allocated in this calculation needs to be properly  
> deallocated. The C++ exception handler does this for us, so replacing  
> it with something else is a major task as we'd need to do our own  
> memory management, or write a garbage collector.
>

Here is a message from March 2004 about the topic.  It compares the  
speed of malloc to new:

     
http://www.octave.org/octave-lists/archive/octave-maintainers.2004/ 
msg00358.html

We might also try std::vector since IIRC it uses malloc internally.

- Paul