> Hi Shaping,

Hi Dirk.

> Shaping wrote:
>>
>> ...   I would like
>> to port OpenSG or OSG to MT or someother multithreaded Smalltalk, if it 
>> exists.  Strongtalk would actually be my first choice, if it had a 
>> graphical debugger and were more stable.  Work on this environ is 
>> progressing but slowly: http://www.strongtalk.org/index.html.
>
> I don't know what is entailed in linking a C++ lib to Smalltalk.

In any descent Smalltalk, you can link a C library (CDECL or STDCALL calling 
convention), and call the C routines like they were methods on a class, 
where the class is an ExternalInterface (VW's term; it goes by other similar 
names in other Smalltalks) representing the C library.  You'd build 
Smalltalk equivalents of the C++ classes, reify as much state as you need on 
the Smalltalk site, and pass these parameters into the C calls.  The guts of 
the C++ methods, of course, would have to be in the C library.  It's a lot 
of work.  I wanted to do this several years ago with Performer, but SGI 
don't give you the source with your license.  I even talked to their lawyer 
about it, and he said that SGI was not interested in a Smalltalk port.  Now 
they're just barely interested at all in anything Performer.

 OpenSG
> has a reflective interface for data, which can make these things 
> 'relatively' easy.

I would like to know more about how this works.

> I know Allen Bierbaum has done one for Python, he might be able to give 
> you some hints if you're interested.

Yes.  Please make him aware of my efforts, if he is not already.

>
>> I do.  SGI have their multipipe SDK programming model, which, as I 
>> currently understand it, uses callbacks into your program(s) running in 
>> different processes/threads, in order to share the rendering load.  My 
>> app could benefit from this, but I don't know yet how I want to partition 
>> the rendering.
>
> OpenSG does that, too, in some contexts. Especially when it comes to using 
> a cluster of machines to split the load, OpenSG makes things fairly simple 
> compared to other solutions. I didn't know that was an issue for you.

It is not so much an issue across a cluster (but could becomes an issue 
latter).  I want to be able to use all of the cores in the new octacore 
machine I'm getting soon.
>
>> How well does it work?  Are there good examples of anyone using OpenSG's 
>> multithreading?
>
> Very well, of course. ;) There are some examples, I don't know if I would 
> call them good, but they show the pieces and this list is very fast in 
> responding to problems if they show up.

I'll look for links to the examples.

>
>> That's OK.  I don't think I would ever want to do that, anyway.
>
> Well, that depends, I guess. If the process of creating your geometry to 
> display is not very fast (wich it might not be if you do it in Smalltalk), 
> it can be useful to prepare the next iteration's geometry while the 
> current one is being drawn.

Oh, I'm thinking then that this is just a second (or nth) scene graph. 
Don't we want to be able to do this in OpenSG?

>
>> I spoke with an SGI engineer today.  He said essentially that Performer 
>> is being supported, but not being advanced actively.  The guys who budded 
>> off OpenSG and OSG have greatly diminished the market for Performer.  Too 
>> much open-source prevents Performer from competing, but I get the 
>> impression that Performer is more finished and ready for serious project 
>> work than OpenSG and OSG.  Is this correct?
>
> it is more finished in the sense that it has pretty good documentation 
> (professional Technical Writers are really helpful for that). Both OpenSG 
> and OSG have decent Tutorials, but not on par with performers 
> documentation. Beyond that, technically, I don't think Performer is better 
> than either one.

OK, good to know.  I still have lots of reading on the site to do.  If I 
talk to the OSG guys, what will they say in defence of OSG?  If there is not 
an earth-shaking difference between OSG and OpenSG, then why don't the two 
of you get together, combine efforts, and discard one of the names (or make 
a new one)?

>
>> I read through these last night.  Most of the differences seem to relate 
>> to supported file-formats.
>
> Those are easy to write, so they tend to get long, yes. ;)
>
>> Someone please tell me all of the things they want very badly to do with 
>> OpenSG, but cannot yet do because OpenSG doesn't support the needed 
>> feature.
>
> I can't really think of anything,

Good to hear, but what about the multiple, concurrently constructed SGs?

My model is a space-and-time pig when it comes to computing all of the 
graphical elements that will be displayed for a large version of it (limited 
only by system memory and your willingness to wait).  I don't have much need 
yet for textures.  The graphical model is all geometry and a little color.

 but then again, I'm the one who
> decides what goes in there, so if something is bad I fix it (or get 
> somebody else to fix it).
>
> Hope it helps

Yes, thanks.


Shaping


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Opensg-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensg-users

Reply via email to