kicad-developers team mailing list archive

[Alex G <mr.nuke.me@xxxxxxxxx>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 09/06/2010 04:58 PM, Lorenzo Marcantonio wrote:
> On Mon, 6 Sep 2010, Alex G wrote:
> 
>> I've actually been surprised by Intel's support in the linux arena. It
>> works out of the box. And yes, my OpenGL code runs on Intel without a
>> hitch (performance not counted).
> 
> Intel doesn't do high perf 3d cards, so it doesn't surprise me :D
> 
It's enough for many scientific visualization task. It even has VBO
support, which is insanely fast. That surprised me.

>> And also, it just happens that the 3D viewer only works on nvidia cards
>> (everything else I tried segfaults, and other than using wxWidgets, I
>> can;t fing the culprit in the kicad code). :-P
> 
> I had the same problem; no idea about where it's the problem (but
> blender works fine on intels... maybe that's because it's 100% opengl).
> 
I'm beginning more and more to find freeglut the most stable and bugfree
windowing alternative. It suck though that it has no threading support,
and that just kills portability when wanting to have multiple windows
with different contexts.

Anyway, there are two bug reports on the kicad issue:
https://bugzilla.redhat.com/show_bug.cgi?id=592047
http://bugs.freedesktop.org/show_bug.cgi?id=29833

The former prompted me to
try
{
    me->readSource(kicad->3dviewer);
}
catch(errorNoCulpritFound)
{
    me->giveUp();
}

>> An old GF2MX400 could handle tons of 3D with millions of polygons.
>> Everything today is more powerful than that, so there's no reason to shy
>> on the quality of the models.
> 
> Be warned that consumer cards may have humungous fill rate but may lack
> in geometry processing... that's why in the old agp days my firegl
> pumped geometry *way* faster than even quadros (it had a dedicated
> geometry engine AND took two slots :D). Modern gpus are actually more
> like general purpose processors so things have changed (look at what
> CUDA does these days...)
> 
Been there, done that. I know all about the internals of GPUs. Dual
9800GT user here, 0.5TFLOP/s with CUDA on software that I wrote. It
actually beats the fastest cluster at my university.

> The real question is... what are complex model for? Eye candy or what?:D
> For most *engineering* processes (i.e. mechanical integration)
> a bounding box or something equally plausible is more than sufficient.
> Connectors could be an exception for their 'live side'.
> 
But you have to also look at the bleeding edgeness of kicad. A bounding
box looks like crap. One reason why I love Oyvind's library is the
vividness of the 3D models. One could design a board for a customer, and
send 3d screenshots of the board. Would that not be appealing?

For engineering, an orthographic view for the 3D viewer is a must. I
will try and patch that once I get libiges out of my way. Hopefully,
I'll learn enough about the internals of kicad to be of some real use.

Alex
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/
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=bqnv
-----END PGP SIGNATURE-----