Re: run-octave on the cygwin

[Tatsuro MATSUOKA]

Hello

--- Marco Atzeri <address@hidden> wrote:

> 
> I run the comparison with your version 
> and mine is 3-4 time slower, as expected.
> 
> I also re-upload the binary for a minor issue. 
> (I moved the libexec dir in the binary in the 
> wrong way instead of rebuilding it)
> 
Thank you for your information.
TestOregoB is a hard test because it requires iterative evaluation for the 
function to solve ODE and
the oregonator ODE is very stiff.
I think that 3-4 time slower is no so slow as exected.  
For the 2.1.73, sometime 5-6 times slower than octave build with gcc-3.2. 
(gcc-3.2 on cygwin handled dwarf2 exception.)

The improvement for speed perhaps relies on the improvement of the cygwin1.dll.
I recommend that you will go ahead further.
For usual use it is the trage off the fast but bug and relatively slow and 
stable.

If one write the fuction by C++ and use mkoctfile, the slowness is not a 
problem.

However for loop slowness remains the problem.
I think that you write such causion for use your octave and recommend to write 
code as vertorized as
possible.  (Of couse it is always true for octave on all platforms.)  

Perhaps you know, the jwe uploaded the pre-release code to maintainer.
I have tested it as your suggestion of run-octave.
It was very useful for me.
New version can be built without essential problems.

Thanks

Tatsuro

  

Regards

Tatsuro

 

--------------------------------------
GANBARE! NIPPON! Win your ticket to Olympic Games 2008.
http://pr.mail.yahoo.co.jp/ganbare-nippon/