On 24 May 2010, at 20:12, Thomas Weber wrote:

> On Mon, May 24, 2010 at 06:06:10PM +0200, c. wrote:
>>
>> On 24 May 2010, at 10:39, Thomas Weber wrote:
>>
>>> Ignoring the bug report,
>> I do not want the bug report to remain ignored, I am willing to help
>> find a fix, just not this way.
>
> I meant this in the sense of "even without the bug report". I guess  
> it's
> fair to say that this is not a real bug in the ocs package.
>
>>> the issue of start-up time for Octave remains.
>> I don't think there is any measurable performance hit due to the more
>> directories added to the path, can you show any numbers supporting
>> this?
>
> With quite some octave-forge packages installed (actually, all that  
> are
> in Debian currently; v.m contains just an exit; command);
>
> $ time octave v.m
> <snip>
>
> real    0m2.820s
> user    0m2.168s
> sys     0m0.112s
>
> without any packages:
> real    0m0.423s
> user    0m0.236s
> sys     0m0.048s
>
> So, the difference is measurable. However, most people will probably
> have longer sessions, where start-up time isn't a big issue.

I tried the same test on my system and the number I see are consistent  
with yours.
The only case where a startup time < 3s may matter though is when  
Octave is being invoked many times within a script.
In that case I would say it's better to invoke Octave with the "-qf"  
option and explicitely load only those packages
that are actually needed, e.g. you might want to try this:

time octave -qf --eval "pkg load ocs" v.m

then try putting all files from ocs into one dir and rerun the test.
anyway you should consider that ocs is marked as "Autoload: no"
so its presence should not usually influence startup times.

> Measuring
> the influence of a bigger loadpath isn't as straightforward as startup
> time, though.

I think directories that are not included in the check for changed  
time-stamps should not influence octave's performance, right?
is the path where debian puts the packages within what  
"ignore_function_time_stamp" considers system directories?

> Reconsidering this is obviously a choice. However, as it seems that
> currently only ocs is hit by this, I'm looking for something else.

what about secs1d and secs2d? they also have subdirectories do they  
suffer the same problem?

>>> [1] http://www.pathname.com/fhs/
>> I'll have a look at the link and see if it helps me find out a way
>> around this problem
>
> Hold your horses. This is a distribution-specific problem. Moving  
> all .m
> files into one directory would be an easy solution, but your  
> explanation
> for the split-up is fair.
>
>       Thomas


------------------------------------------------------------------------------

_______________________________________________
Octave-dev mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/octave-dev

Reply via email to