[Sidefx-houdini-list] Curves normals

Olex P hoknamahn at gmail.com
Fri Sep 3 05:39:51 EDT 2010


I think answer is pretty simple - kd-tree gives better performance when it's
built while BVH has better construction time. So probably in general
performance gain is given by faster raytracing than by faster building of
acceleration structures. BVH is better in interactive applications where
construction time is more important.

2010/9/3 Pablo Giménez <pablogipi at gmail.com>

> 2010/8/31 Olex P <hoknamahn at gmail.com>:
> > Hey Pablo,
> >
> > Was it you asking about IPR and fur rendering? Something about IPR
> getting
> > stuck on the kd tree generation stage. Could you try the same scene but
> with
> > BVH (ta-dam we have it now!) instead of kd tree (Mantra ROP, Rendering
> tab,
> > Ray Tracing Accelerator parameter). Construction of BVH should be way
> faster
> > than of kd tree plus it could be less expensive in terms of memory
> > consumption. Maybe it could solve your problem.
> Yes it i much faster, actually makes raytracing with hair usable for me.
> Thanks!
> One thing I have read the docs about vm_octreestyle, but is still not
> clear why BVH is not the degault octree style rather than KD-Tree, in
> other words ,are there adventages of KD-Tree over BVH?
> >
> > 2010/8/31 Pablo Giménez <pablogipi at gmail.com>
> >
> >> 2010/8/28 Szymon Kapeniak <szymon.kapeniak at gmail.com>:
> >> > 2010/8/28 Peter Bowmar <pbowmar at gmail.com>:
> >> >> Indeed, there is a new property that turns on/off shading curves as
> >> >> surfaces. I don't remember what it's called though...
> >> >
> >> > ..."vm_curvesurface" as mentioned previously, but as far as I
> >> > understand Pablo's problem was, that even with this property turned
> >> > on, early build of Mantra H11 didn't create "s" variable for hair (I
> >> > can't verify this though). With H11.0.485 all seems to be fine.
> >> Yes, that was the problem, but now seems to be working in latest daily
> >> built.
> >> >
> >> > Anyhow, this optimization is interesting, I made some test renders
> >> > yesterday comparing H10 and H11 with Marschner, Arno Zinke nad
> >> > Houdini's default textured hair shaders. It seems to work well both
> >> > for rendering speed and memory footprint but also has a huge impact on
> >> > fur quality in a middle to close-up shots. No elliptical hairs shape,
> >> > flat less vivid speculars, less distinctive hairs in a far field
> >> > scenario.
> >> >
> >> >
> >> >
> >> >
> >> >>
> >> >> On 28 August 2010 01:34, Louis Dunlevy <louisdunlevy at gmail.com>
> wrote:
> >> >>> actually I betcha it's just a new default rendering parameter for
> that
> >> >>> optimisation that was mentioned earlier, you can probably switch it
> >> back
> >> >>> to the old method..
> >> >>>
> >> >>> On 08/28/10 00:58, Pablo Giménez wrote:
> >> >>>> Yes definitely seems to be a bug in H11.0.446.9 , yes you got it is
> >> >>>> the current production version!
> >> >>>> Latest daily build work well and creates the s coordinate properly.
> >> >>>> H10 also works fine.
> >> >>>> Just in case somebody else is havign a similar problem.
> >> >>>>
> >> >>>> 2010/8/26 Louis Dunlevy<louisdunlevy at gmail.com>:
> >> >>>>
> >> >>>>> parametric s should not be zero. If it is, I'm thinking it may be
> a
> >> bug
> >> >>>>>
> >> >>>>> On 27/08/10 03:02, Pablo Giménez wrote:
> >> >>>>>
> >> >>>>>> Hi Andrew.
> >> >>>>>> I have done similar things for other stuff, but my problem is
> that
> >> the
> >> >>>>>> s coordinate is constant 0 for all the points in the curve
> geometry.
> >> >>>>>> The t is fine, going from 0 to 1 along the curve lenght, but s is
> >> >>>>>> supposed to be along the curve "cross sections" width, and this
> is
> >> not
> >> >>>>>> happening.
> >> >>>>>> With a proper s coordinate along the width of the curve would be
> >> easy
> >> >>>>>> to fake a rounded look.
> >> >>>>>> cheers
> >> >>>>>>
> >> >>>>>> 2010/8/26 Andrew D Lyons<tstexture at gmail.com>:
> >> >>>>>>
> >> >>>>>>
> >> >>>>>>> I did this recently to generate fur sprites. If you get the
> curve
> >> >>>>>>> tangent and stack some cross products with another version of
> the
> >> >>>>>>> curve with normals hard coded to point towards the camera (or
> some
> >> >>>>>>> cardinal axis) then you should be able to derive a normal that
> is
> >> >>>>>>> always pointing in a general direction - but which follows the
> >> >>>>>>> contours of the curve.
> >> >>>>>>>
> >> >>>>>>> 2010/8/26 Pablo Giménez<pablogipi at gmail.com>:
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>>> Hi.
> >> >>>>>>>> I am rendering some poly lines as curves, using the 'width'
> >> attribute
> >> >>>>>>>> to control every filament width.
> >> >>>>>>>> That's all right.
> >> >>>>>>>> Problem is that this geometry is created at render time and the
> >> >>>>>>>> normals and texture coordinates are a little bit odd.
> >> >>>>>>>> My main problem is that Mantra creates the needed geometry
> using
> >> the
> >> >>>>>>>> width attribute but seems that is only shading the original
> curve
> >> >>>>>>>> attributes.
> >> >>>>>>>> I mean the hair geometry is created extendin the shaded point,
> >> >>>>>>>> perpendicular to the point of view, and the new geo seems to
> use
> >> the
> >> >>>>>>>> same shading as the original point.
> >> >>>>>>>> Is similar to what happens with points rendering, where the
> point
> >> is
> >> >>>>>>>> shaded once, and then all the geometry created at render time
> >> around
> >> >>>>>>>> the origian lpoint use the same shading from the original one.
> >> >>>>>>>> So you don't have proper normals to make for instance a diffuse
> >> >>>>>>>> lighting that gives you some "rounded" look.
> >> >>>>>>>> I have tried the hairNormalVOP, creating my own tangents,
> created
> >> >>>>>>>> texture coordinates with s going along the hair line from 0 to
> 1,
> >> but
> >> >>>>>>>> the normals I got are just pointin towards camera, they are not
> >> doing
> >> >>>>>>>> the "Rounding2 effect that is suppose to do this operator.
> >> >>>>>>>> So is there any way to go away from the planar look of curves
> >> shading,
> >> >>>>>>>> any way to have sense whre you shading point is in the new
> created
> >> >>>>>>>> gemetry so you can for instance fake normals o apply any othe
> >> rtrick.
> >> >>>>>>>> Thanks
> >> >>>>>>>> --
> >> >>>>>>>> Un saludo
> >> >>>>>>>> Best Regards
> >> >>>>>>>> Pablo Giménez
> >> >>>>>>>> _______________________________________________
> >> >>>>>>>> Sidefx-houdini-list mailing list
> >> >>>>>>>> Sidefx-houdini-list at sidefx.com
> >> >>>>>>>>
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >> >>>>>>>>
> >> >>>>>>>>
> >> >>>>>>>>
> >> >>>>>>>
> >> >>>>>>> --
> >> >>>>>>> =======================================
> >> >>>>>>> Andrew D Lyons | Digital Artist | http://www.tstex.com
> >> >>>>>>> =======================================
> >> >>>>>>> _______________________________________________
> >> >>>>>>> Sidefx-houdini-list mailing list
> >> >>>>>>> Sidefx-houdini-list at sidefx.com
> >> >>>>>>>
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>
> >> >>>>>>
> >> >>>>>>
> >> >>>>> _______________________________________________
> >> >>>>> Sidefx-houdini-list mailing list
> >> >>>>> Sidefx-houdini-list at sidefx.com
> >> >>>>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >> >>>>>
> >> >>>>>
> >> >>>>
> >> >>>>
> >> >>>>
> >> >>>
> >> >>> _______________________________________________
> >> >>> Sidefx-houdini-list mailing list
> >> >>> Sidefx-houdini-list at sidefx.com
> >> >>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >> >>>
> >> >>
> >> >>
> >> >>
> >> >> --
> >> >> No, I am not on Facebook.
> >> >> _______________________________________________
> >> >> Sidefx-houdini-list mailing list
> >> >> Sidefx-houdini-list at sidefx.com
> >> >> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >> >>
> >> > _______________________________________________
> >> > Sidefx-houdini-list mailing list
> >> > Sidefx-houdini-list at sidefx.com
> >> > https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >> >
> >>
> >>
> >>
> >> --
> >> Un saludo
> >> Best Regards
> >> Pablo Giménez
> >> _______________________________________________
> >> Sidefx-houdini-list mailing list
> >> Sidefx-houdini-list at sidefx.com
> >> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >>
> > _______________________________________________
> > Sidefx-houdini-list mailing list
> > Sidefx-houdini-list at sidefx.com
> > https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >
>
>
>
> --
> Un saludo
> Best Regards
> Pablo Giménez
> _______________________________________________
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>



More information about the Sidefx-houdini-list mailing list