halfdan
Isn't everyone jumping on the VFX Reference Platform bandwagon anyway?

Indeed we are.

VFX Platform CY2014 (Current)
gcc 4.1.2

*ahem*

Any chance you want to (or don't want to but still can) start building with GCC 4.1 again?

We have a rather massive codebase built in 4.1 which we'd rather not rebuild using 4.4+ just yet.

To add a touch more detail to Guillaume's post, we tried the simple star SOP example. Built fine, crashed when used.

Apologies if this was answered many moons ago when the decision was made (my searching skills failed me), but why were the GCC 4.1 builds dropped? Was there a technical issue/limitation, or was it “only” an administrative annoyance?

After a long hiatus in building our tech in/for Houdini, we recently discovered to our horror that building plugins using GCC4.1 works, but they crash when actually used (in the GCC4.4-built Houdini 13.5).

Cheers!

- Rob

edward
Hi Rob. Try this?

http://www.itworld.com/software/306674/disable-visual-effects-make-ubuntu-1210-faster [itworld.com]

I did, yeah (I was reading that exact page this morning, in fact). It definitely affected the startup splash screen, but didn't seem to affect the pop-up menu.

I'm going to try installing 12.5 here (on Centos6) to see what that's like.

I'm not sure which the culprit is (Houdini or Ubuntu), but when I hit <tab> to bring up the list of nodes I can use, it semi-quickly fades and scales into place rather than just popping-up. By itself this is okay, but as soon as I start typing and the list gets filtered, the original window/list scales out and then a new one scales in.

It's just slow enough to be really annoying

I didn't see anything in the Houdini prefs about this, so I'm guessing it might be an Ubuntu thing. But I'm polling for advice.

Thanks!

- Rob

THANK YOU!!! I was going insane trying to figure out why Houdini would keep hanging/crashing on me after ~30sec.

twod
Which graphics card are you using? Have you updated your graphics driver to a recent driver?

FWIW, I got the exact same thing out-of-the-box with Ubuntu 12.10. The issue went away when I switched to the NVIDIA driver.

Just ran into this here too. In case it helps anyone in the future, the issue for me was definitely related to python & associated libs.

Final solution:

source houdini_setup
setenv PYTHON_HOME where-my-python-actually-is
setenv LD_LIBRARY_PATH something-awesome:${HFS}/python/lib
houdini

So let's say I've got an awesome particle simulation that I can comfortably preview as 1000-points but want to render as 1000000-points.

No problem, I can go to the emitter nodes and change the number back and forth manually depending on if I'm tweaking the sim or doing a render. But this is very manual and prone to error.

Or I could set up two takes, one for preview and one for render, and have the emitter parameters set to take-specific values. But if the values are part of a complicated expression that I might want to change, then I have to remember to change the expression in both takes, or split the expression into the common bit and a take-specific multiplier which is stored in a different parameter. Still feels kind of messy.

How do you typically handle this? Personally I'd love to see (maybe it exists?) a command to let me query if a cook was due to a display or render call. This way my particle emission could be controlled like:

if(cookReason()=ISPLAY,1000,1000000)

Sorry, didn't mean to strike a nerve.

On the plus side, it's great to be “back” (even if it's just in my spare time) and see the same names on the forum as were there when I was at C.O.R.E.

Mario Marengo

rjpieke
Any clue why:
1) Ce isn't exposed in the material builder?

Ce was left as a user export during its inception in version 10… and has gradually become a more bona-fide (that is, “expected”) export as time's gone by. At this point, I agree there's probably a good argument for making it a standard shader export (at the same level of Cf, Af, F, etc.)… but there might be some reason that's holding that back; I dunno. <shrug>

Gah, I was hoping to dive straight into H11, but I suspect I've got some H10 (and possibly H9) reading still to do first. I've definitely got “what's the philosophy behind THAT decision?” questions arising (the wealth of rendering parameters it seems I have to keep adding to every object, for example) that are probably explained to death in the archives of this forum

Thanks Mario! That did the trick beautifully!

Any clue why:
1) Ce isn't exposed in the material builder?
2) Auto-layout doesn't work in VOPs?

How's Toronto? I _still_ haven't managed to seize the opportunity to sneak back for some Burrito Boyz

Attachment of non-working scene

Another H11 feature I can't quite get working yet. The way I read the docs, I should create some geo, attach a material, create a geometry light, and the light will use that material.

Tried this and got a black frame. Enabling “render light geometry” showed my geometry with the material I'd chosen, but with no effective illumination of the scene. Removing the material results in a pure white geometry with an even scene illumination.

ikerr
Try rendering using ray-tracing

Aha … that was the missing piece. Did I totally miss that in the documentation, or do you just have the magic mojo?

ikerr
I've attached an example file showing how.

Which build are you using? I tried loading your scene into 11.0.446.9 (the production build) and got a black frame.

Poking through the “What's New in H11”, I read about vm_volumeuniform. So I added that property to an object in my scene and clicked render. The geometry rendered exactly the same. Tried adding the “Basic Smoke” material, and it didn't render at all.

On an unrelated note, between the odforce forums, these forums, and the sidefx mailing list, where do people feel there's the most activity these days?

eetu
Can repro, yes.
My personal guess would
be that this just the “normal” floating point accuracy issue. Floating
point numbers are ill suited for storing exact values, and you should probably
test against a small epsilon value instead of absolute equality. In this case e.g. diff < 0.000001
eetu.

Generally floats aren't good for specific values, but if your number can be expressed as the sum of powers of 2 (in this case I'm storing 1, which is 2^0), they _should_ be perfect.

Any thoughts? Can anyone at least reproduce this?