Text archives Help

--- Begin Message ---

• From: "Brian Budge" <brian.budge@gmail.com>
• To: "Thiago Ize" <thiago@cs.utah.edu>
• Subject: Re: [Manta] Re: manta
• Date: Tue, 25 Dec 2007 16:17:27 -0700
• Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:in-reply-to:mime-version:content-type:references; b=eLAyekr27faf8KHXqVzgG9RGAOv8OeeAtMwNn9a1/888h0/Ro/mmGhMTdXYBdsL8rm3u5jCEQOZeqnQ/aJROt7gjtfgHGzsHFXbKryUCHoi4FHxCMN5xfbw5gthBPhWZq3FME3IpjMr9B0LfM/GrpvD/Hg7AiZRV7E4xXErrXIg=

Hi Thiago -

 

Thanks for the pointers.  It seems as though all is well with how I'm running these models through.  I changed the approximate flag in the ccmake config, and it is a might bit faster for rendering and quite a lot slower for building.  The scene from the vantage listed below runs at about 2.8 fps, which sounds about right, considering my Core2 is running at 2.0 GHz.

It is still a decent amount faster than my code... I get about 1.6 fps to get roughly the same image.

 

Thanks again for the help.  I wonder if I can ask one more question.  Is it possible to assign a BRDF friendly material to a model loaded with the triangleViewer?  It would be nice to run a model through with, say, a lambertian shader, so that I can get normal looking path-traced images.  Would it be hard to do this?

 

Thanks!  You've all been very helpful.

 

  Brian

 

On Dec 24, 2007 3:03 PM, Thiago Ize <thiago@cs.utah.edu> wrote:

The following:
bin/manta -scene "lib/libscene_triangleSceneViewer.dylib(-DynBVH -model
/Users/thiago/data/models/happy_vrip.ply -triangleType Wald_tri)"
-imagetraverser "tiled (-square )" -camera "pinhole( -eye 0.00671022
0.145579 0.191262 -lookat -0.00543931 0.148768 -0.00668995 -up
-0.0247273 0.999539 0.0176212 -hfov 60 -vfov 60 -normalizeRays
-createCornerRays)" -res 1280x720

Gives me 3fps on my 2.1GHz core 2 duo.  That's the standard 1M tri
buddha (if you're usng linux, you'll need to change .dylib to .so). In
order to get a more direct comparison to published DynBVH results, you
need to run ccmake and set  MANTA_USE_DYNBVH_APPROXIMATE to OFF (this
uses a slower but better build and gives 3.4fps) and
MANTA_USE_DYNBVH_PORTS  to ON (ports is a direct port from the code used
for our papers, but it's buggy in manta if you start doing secondary
rays). With these changes I get 4.8fps.

A quick way to check if you have sse enabled is to run ccmake and check
if MANTA_SSE is on. Also, are you running the code in Release mode? When
you compile with make VERBOSE=1 check to see whether you see all the
optimization flags as well as the MANTA_SSE define being sent to the
compiler.

Also, running bin/manta with no options (all default) gives almost 20fps
on my system.

Oh one final thing. I believe there is a bug in DynBVH that is keeping
it from displaying shadows (do you see shadows?) That would also help
explain a little why the kdtree is slower (I'm getting 1fps with 64 rays
and 1.6 with 16 rays per packet, but a least it's actually doing
shadows). You can disable shadows by adding -shadows noshadows

Thiago

Brian Budge wrote:
> Hi all -
>
> The timings I listed were only ray casting + simple shading based on
> the normal (At least I'm assuming that's the reason for the green and
> purple hues).
>
> I was running with a single ray tracing thread on my Sony Vaio laptop,
> which has a Core2 Duo 7200.
>
> I noticed the timings in the paper were for 512x512 images and I was
> running 1280x720, so that may be part of the discrepancy.  Maybe
> somehow the optimal packet sizes aren't being used?
>
> The buddha scene I listed before isn't the one in the Stanford
> repository.  I can try some tests with that one (standard 1M scene):
>
> DynBVH: 0.3 seconds/frame at 512x512, 0.7 seconds/frame at 1280x720
> Manta KD: 1.5 seconds/frame at 512x512, 2.6 seconds/frame at 1280x720
> My KD: 0.55 seconds/frame at 512x512, 1.5 seconds/frame at 1280x720
>
> I still think I may not have SSE packets working properly for Manta.
> How can I check that it is working properly?  BTW, these measurements
> are very eye-point dependent.  For example with the DynBVH looking
> down at the Buddha head I get over 10 frames per second, but facing
> from the front I get 3 fps.
>
> Merry Xmas!
>   Brian
>
>
>
> On Dec 23, 2007 10:34 PM, Ingo Wald <ingowald@gmail.com

> <mailto:ingowald@gmail.com>> wrote:
>
>     Given that it takes 1+ sec / frame, that strongly suggests that either
>     something very, very weird is going on here, or that Brian indeed
>     uses
>     multiple secondary rays. In that case I'd not be surprised if
>     performance tanks; we've never played around with that, and many
>     optimizations might actually be disabled for secondary rays, anyway
>     (@brian: at least in the original dynrt codebase, only certain
>     types of
>     rays actually used all the optimization techniques -- I'm not sure
>     what
>     has been added since that code has been pulled over into manta).
>
>     Merry Christmas ;-)
>
>     Ingo
>
>     Thiago Ize wrote:
>     > Brian,
>     > Are you using manta for shooting just primary rays or are you
>     doing a
>     > full blown path tracer? If it's a path tracer with lots of bounces,
>     > then I wouldn't be at all surprised that your single ray
>     acceleration
>     > structure would perform better since it wouldn't have all the
>     overhead
>     > inherent in the DynBVH and KDTree acceleration structures which are
>     > currently only designed for packets. If you are just raycasting,
>     then
>     > something is very wrong, since you're at least an order of magnitude
>     > too slow.
>     >
>     > Thiago
>     >
>     > PS: Why are you not using the standard 1M tri buddha?
>     >
>     > Brian Budge wrote:
>     >> Hi Ingo -
>     >>
>     >> I just ran a couple of experiments through again.  It looks as
>     though
>     >> the DynBVH is definitely faster than both kd-tree implementations.
>     >> Running the 1.5 million triangle buddha scene, I get the following
>     >> timings (at 1280x720):
>     >>
>     >> DynBVH: 1.1 seconds/frame
>     >> Manta KD: 1.7 seconds/frame
>     >> My KD: 1.4 seconds/frame
>     >>
>     >> I'm a total Manta newbie, so it may be possible that I don't have
>     >> packet tracing turned on somehow?  I think it's on by default?
>     >>
>     >> I guess there is an untold story here too.  If I run smaller
>     scenes
>     >> (say 7,500 triangles), the timings are quite a bit different:
>     >>
>     >> DynBVH: 0.33 seconds/frame
>     >> Manta KD: 0.3 seconds/frame
>     >> My KD: 1.2 seconds/frame
>     >>
>     >> I can't say I understand this too well, but I would venture a
>     guess
>     >> that the packets stay coherent much longer with the smaller scene.
>     >> Either that or I am doing something stupid with small scenes that
>     >> doesn't show up with larger scenes.
>     >>
>     >> Any insights?
>     >>
>     >> BTW, I am using a simple traversal scheme very similar to the
>     one in
>     >> your thesis.
>     >>
>     >> Thanks,
>     >>   Brian
>     >>
>     >> On Dec 23, 2007 1:12 AM, Ingo Wald < ingowald@gmail.com
>     <mailto: ingowald@gmail.com>

>     >> <mailto:ingowald@gmail.com <mailto:ingowald@gmail.com>>> wrote:
>     >>
>     >>
>     >>     >
>     >>     >     > for the viewer?  My own kd-tree ray tracer without
>     sse is
>     >>     about the
>     >>     >     > same
>     >>     >     > speed as tte DynBVH that you guys have, but I would
>     guess
>     >>     that your
>     >>     >     > kd-tree
>     >>     >     > is a good deal faster.
>     >>     >
>     >>     Hi Brian,
>     >>
>     >>     I actually doubt it is ;-).
>     >>     I've written the SSE'd kd-tree traverser and kd-SAH builder
>     >> (assuming
>     >>     the  one in manta is the one I wrote for Thiago's project...),
>     >> but it
>     >>     certainly isn't faster than DynBVH (how could it !?).
>     >>     I'm also a bit surprised that your _single-ray_ kd-tree is as
>     >> fast as
>     >>     DynBVH, which would be _very_ strange indeed (in theory,
>     "perf(SSE
>     >>     packet/frustum kd-tree) > perf(single-ray kd-tree)" and
>     >>     "perf(dynbvh) >
>     >>     perf(SSE kd-tree)", so your message confuses me a bit.
>     >>
>     >>     Can you elaborate a bit on
>     >>     a) what traverser you actually use, and
>     >>     b) what performance you're getting with dynbvh ? anything close
>     >> to the
>     >>     numbers we reported in the paper ?
>     >>
>     >>     Ingo
>     >>
>     >>
>     >
>     >
>
>

--- End Message ---

• [Fwd: Re: [Manta] Re: manta], Thiago Ize, 12/25/2007
  • Re: [Fwd: Re: [Manta] Re: manta], Steven G. Parker, 12/25/2007

Archive powered by MHonArc 2.6.16.