Text archives Help

---

It appears that the -createCornerRays option, added in r1635 causes a performance hit with the default packet size and breaks camera manipulation in that code. It's also possible that this problem only occurs on osx/gcc. Thiago is taking a closer look.

On the bright side this code produces no warnings when compiled ;-)

Abe

On Aug 15, 2007, at 5:57 PM, Solomon Boulos wrote:

With my settings (64 rays as RayPacket::MaxSize and 64 as Fragment::MaxSize) I don't see any change (though I'm not comparing to runwxmanta). Perhaps the recent camera change doesn't work for the default settings (32 as maxsize)? What are your settings?

Also what kind of manipulation did you perform to the camera to get it to do produce the ambiguous up directions? I couldn't trigger this (even with Maya).

On Aug 15, 2007, at 5:53 PM, Abe Stephens wrote:

I just noticed that the performance of bin/manta is considerably less than it was before SIGGRAPH. On one laptop core bin/manta runs at 7.6 fps while runwxmanta.py runs at 18.0 fps on one core. On two cores bin/manta drops to 3.6 fps while the python interface goes to ~34 fps.

Also, manipulating the camera in bin/manta appears to cause a whole bunch of nan errors making interaction impossible.

/Users/abe/Manta/Model/Cameras/PinholeCamera.cc line: 133 Ambiguous up direciton... lookat = nan nan nan, eye = nan nan nan, direciton = nan nan nan, n = nan nan nan, up = nan nan

In short, bin/manta appears to be broken for the default scene.

Anyone know what is going on?

Abe

---

• [Manta] bin/manta broken., Abe Stephens, 08/15/2007
  • Re: [Manta] bin/manta broken., Solomon Boulos, 08/15/2007
    • Re: [Manta] bin/manta broken., Abe Stephens, 08/15/2007

---

Archive powered by MHonArc 2.6.16.