pbrt v2 alpha is now available

[Matt Pharr]

We're happy to announce that a first alpha release of the second version of 

pbrt is now available for the brave and adventurous.  This version of the

source code corresponds to the system that will be described in the second

edition of the book "Physically Based Rendering: From Theory To

Implementation", which is expected to be available at SIGGRAPH 2010.

The source code is available from github: http://github.com/mmp/pbrt-v2.

The easiest way to download the code and track changes is to use the git

source code management system:

% git clone git://github.com/mmp/pbrt-v2.git pbrt-v2

But you can also click the "download" button on that page to download a 

tar or zip file of the latest version of the code.

We are making the source code available before publication of the book in

the hopes that it will be useful to developers and users.  We in

particular hope that we will receive some feedback on the changes to the

system: comments on better implementation approaches, and performance or

portability shortcomings would be highly welcomed!  If you do have any

feedback, please either send it to the pbrt mailing list, or to

aut...@pbrt.org.

The remainder of README file (with much more information about what

has changed, etc.) is below.

Thanks,

-matt

We will release updated versions of the system frequently; the latest

version of the code will always be available at the github page 

http://github.com/mmp/pbrt-v2.  The best way to track changes to the system 

is to use the git source code management system, although tar files and 

zip files are also available for download from the github page.

The remainder of this document will summarize the major changes to the

system.  See the file README_BUILDING.txt for information about how to

compile the system.

--- Incompatibilities ---

Many existing pbrt scene files will work unmodified with the second version

of the system.  The most significant user-visible change is that we have

taken this opportunity to fix the long-standing bug where LookAt

inadvertently flipped the handed-ness of the coordinate system

(http://pbrt.org/bugtracker/view.php?id=5).  You may find that scene files

that use LookAt now render flipped images or otherwise have problems with

the camera positioning.  Add "Scale -1 1 1" to the top of any scene

description files with this problem to return to the previous camera

position.

pbrt no longer has a plugin implementation based on runtime loading of DSOs

or DLLs.  All implementations of shapes, lights, etc., are just statically

compiled into the pbrt binary.  (We found that the complexity of supporting

this, both in cross-platform headaches, user issues with plugins not being

found, and so forth wasn't worth the limited benefits.)  The

PBRT_SEARCHPATH environment variable is no longer used, and the SearchPath

directive has been removed from the input file syntax.

We have removed the "shinymetal" material, replacing it with a physically

based "metal" material.  (Described further in the new features section

below.)

--- New Features ---

-- General --

pbrt is now multithreaded; performance should scale well with increasing

numbers of CPU cores.  We are particularly interested to hear any

experience running the system with >4 cores as well as any insight gained

from digging into any scalability bottlenecks.  The system attempts to

automatically determine how many CPU cores are present in the system, but

the PBRT_NTHREADS environment variable can be used to override this.

OpenEXR is no longer required to build the system (but it is highly

recommended).  pbrt now includes code to read and write TGA format files;

support for that file format is thus always available.  If pbrt is compiled

with PBRT_HAS_OPENEXR #defined, then OpenEXR files can be used as well.

(With the file suffix determining which format is used.) 

An accelerator based on bounding volume hierarchies has been added

(accelerators/bvh*).  This accelerator is now the default; it spends

substantially less time than the kd-tree accelerator in hierarchy

construction, while providing nearly equal performance. 

pbrt now supports full spectral rendering as a compile-time option; to

enable it, change the "typedef RGBSpectrum Spectrum" in core/pbrt.h to

"typedef SampledSpectrum Spectrum".  The number of spectral samples taken

(30 by default, leading to 10nm spacing), can be changed in the

core/spectrum.h file.

Animation is now supported via animated transformations for cameras and

shapes in the scene.  (But not light sources, however.)  See the included

example files scenes/anim-killeroos-moving.pbrt and

scenes/anim-moving-reflection.pbrt.  It is likely that the method for

specifying animated transformations in input files will change in order to

be more similar to that used by lux render.

A rudimentary adaptive sampler is included, see samplers/adaptive.*.

-- Integrators --

The 'instant global illumination', 'extended photon map', and 'extended

infinite area light source' implementations from the author-supplied

plugins at http://pbrt.org/downloads.php are now standard; the previous

photon map and infinite area light source implementations described in the

first version of the book have been removed.

The irradiance cache implementation has been substantially improved,

following many of the ideas from the Tabellion and Lamorlette paper from

SIGGRAPH 2004.

A subsurface scattering integrator based on the Jensen and Buhler 2002

SIGGRAPH paper has been added (integrators/dipolesubsurface.*).  Two new

materials for translucent materials have been added, "kdsubsurface" and

"subsurface".  The former implements Jensen and Buhler's more intuitive

approach for setting subsurface scattering parameters.  The latter also

incorporates measured scattering parameters from a number of recent papers;

many translucent material types are available by name.  (See the source

code for details.)

An implementation of Metropolis light transport has been added (based on

the Kelemen et al 2002 Eurographics paper).  This is implemented as a new

"renderer", rather than for example a surface integrator.  See the included

example scene file scenes/metal.pbrt.

Support for a variety of forms of precomputed radiance transport has been

added.  A new renderer that computes radiance light probes in spherical

harmonics on a grid has been added (renderers/createprobes.*), and a

surface integrator that uses them is available in integrators/useprobes.*.

The diffuse precomputed radiance transfer (PRT) method of Sloan et al's 2002

SIGGRAPH paper is implemented in integrators/diffuseprt.* and a technique

for glossy PRT is in integrators/glossyprt.*.  Note that in general, one

would use pbrt to do the precomputation part of PRT and then use the

results in a real-time rendering system.  Here we have also implemented the

code that uses the results within pbrt for pedagogical purposes (and so

that we don't need to include a real-time renderer with the book!).

-- Materials --

A new 'metal' material has been added; it supports setting the spectral

index of refraction and extinction coefficients via measured data from real

metals.  See the example file scenes/metal.pbrt.  A large number of

spectral measurements are available in the scenes/spds/metals/ directory.

Support for measured BRDF data is now available via the "measured"

material.  Two BRDF representations and file formats are supported.  First

is the binary file format used by the MERL BRDF database.

(http://merl.com/brdf).  For arbitrary scattered BRDF measurements, a

general text file format is supported; see the comments in

scenes/brdfs/acryl_blue.brdf.

--- Example Scenes ---

A small number of example scenes that demonstrate some of the new features

are provided in the scenes/ directory.

anim-killeroos-moving.pbrt, anim-moving-reflection.pbrt: demonstrates

motion blur features.

bunny.pbrt: measured BRDFs

metal.pbrt: Metropolis light transport, measured BRDFs

prt-teapot.pbrt: precomputed radiance transfer

ss-envmap.pbrt: subsurface scattering

[Joo-Haeng Lee]

It's great!

I got no error while rendering all sample scenes in my iMac, after building pbrt-v2 with no pain.

I am going to try in two more machines: (1) Vista 64 with VS2005, and (2) OSX both on MacPro.

Thanks.

/Joo-Haeng

[Rachit Agrawal]

It builds very smoothly.

There are some queries that I have:

1. There are some textures that are removed like shinymetal. How to I
render the scenes using these materials? Will the material "metal" do?
2. How can I specify the number of threads I want to spawn?
3. There seem to be problem with almost all my old scene files. Can
they be fixed? These are the following errors:

buddha.pbrt
---------------

geometry/happy.pbrt(1631379): Warning: Material "bluepaint" unknown.
geometry/happy.pbrt(1631379): Error: Unable to create material
"bluepaint"

dof-dragons.pbrt
---------------------

geometry/dragon.pbrt(1309069): Warning: Material "bluepaint" unknown.
geometry/dragon.pbrt(1309069): Error: Unable to create material
"bluepaint"

geometry/dragon.pbrt(1309069): Warning: Material "clay" unknown.
geometry/dragon.pbrt(1309069): Error: Unable to create material "clay"

dragon.pbrt
--------------

dragon.pbrt(81): Warning: Parameter "directphotons" not used
dragon.pbrt(81): Warning: Parameter "directwithphotons" not used

Shooting photons: [+++++++++++++++++++++++++++++++++ ]
(224.2s|61.3s) terminate called after throwing an instance of
'std::bad_alloc'
what(): std::bad_alloc
Aborted

tt-tire.pbrt
------------
geometry/tt.pbrt(885986): Warning: Parameter "Ks" not used
geometry/tt.pbrt(885986): Warning: Parameter "roughness" not used
geometry/tt.pbrt(885990): Error: Expected numeric parameter value for
parameter "roughness" with type "float". Ignoring.
geometry/tt.pbrt(885990): Error: Expected numeric parameter value for
parameter "bumpmap" with type "float". Ignoring.
geometry/tt.pbrt(958231): Warning: Material "shinymetal" unknown.
geometry/tt.pbrt(958231): Warning: Parameter "Ks" not used
geometry/tt.pbrt(958231): Warning: Parameter "roughness" not used
geometry/tt.pbrt(958231): Error: Unable to create material
"shinymetal"

tt.pbrt
-------
geometry/tt.pbrt(238175): Warning: Parameter "Ks" not used
geometry/tt.pbrt(261795): Error: Expected numeric parameter value for
parameter "uroughness" with type "float". Ignoring.
geometry/tt.pbrt(270096): Error: Expected numeric parameter value for
parameter "uroughness" with type "float". Ignoring.
geometry/tt.pbrt(276304): Error: Expected numeric parameter value for
parameter "uroughness" with type "float". Ignoring.
geometry/tt.pbrt(278858): Error: Expected numeric parameter value for
parameter "uroughness" with type "float". Ignoring.
geometry/tt.pbrt(281002): Error: Expected numeric parameter value for
parameter "bumpmap" with type "float". Ignoring.
geometry/tt.pbrt(282007): Error: Expected numeric parameter value for
parameter "bumpmap" with type "float". Ignoring

Can any help me how to rectify these errors? Thanks

On Oct 8, 8:22 am, Matt Pharr <matt.ph...@gmail.com> wrote:
> We're happy to announce that a first alpha release of the second  
> version of
> pbrt is now available for the brave and adventurous.  This version of  
> the
> source code corresponds to the system that will be described in the  
> second
> edition of the book "Physically Based Rendering: From Theory To
> Implementation", which is expected to be available at SIGGRAPH 2010.
>
> The source code is available from github:http://github.com/mmp/pbrt-v2.
> The easiest way to download the code and track changes is to use the git
> source code management system:
>
> % git clone git://github.com/mmp/pbrt-v2.git pbrt-v2
>
> But you can also click the "download" button on that page to download a
> tar or zip file of the latest version of the code.
>
> We are making the source code available before publication of the book  
> in
> the hopes that it will be useful to developers and users.  We in
> particular hope that we will receive some feedback on the changes to the
> system: comments on better implementation approaches, and performance or
> portability shortcomings would be highly welcomed!  If you do have any
> feedback, please either send it to the pbrt mailing list, or to

> auth...@pbrt.org.

>
> The remainder of README file (with much more information about what
> has changed, etc.) is below.
>
> Thanks,
> -matt
>
> We will release updated versions of the system frequently; the latest

> version of the code will always be available at the github pagehttp://github.com/mmp/pbrt-v2.  The best way to track changes to the  

> plugins athttp://pbrt.org/downloads.phpare now standard; the previous

[John Danks]

On Fri, Oct 16, 2009 at 5:11 AM, Rachit Agrawal <rachit.k...@gmail.com> wrote:

It builds very smoothly.

There are some queries that I have:

1. There are some textures that are removed like shinymetal. How to I
render the scenes using these materials? Will the material "metal" do?

Metal should work, but you'll have to look at the parameters for both the old material and the new one to get a good match.
 

2. How can I specify the number of threads I want to spawn?

PBRT v2 makes 1 thread for each processor core it detects. You can force it create a specific number of threads by defining the PBRT_NTHREADS environment variable.
 

3. There seem to be problem with almost all my old scene files. Can
they be fixed? These are the following errors:

I can't speak for all your scene errors, but...

dragon.pbrt
--------------

dragon.pbrt(81): Warning: Parameter "directphotons" not used
dragon.pbrt(81): Warning: Parameter "directwithphotons" not used

Shooting photons: [+++++++++++++++++++++++++++++++++         ]
(224.2s|61.3s)  terminate called after throwing an instance of
'std::bad_alloc'
 what():  std::bad_alloc
Aborted

Dragon is a _very_ large model. You are most likely running out of memory.

[Matt Pharr]

(Second reply to the original email coming in a few minutes as well.)

On Oct 16, 2009, at 10:33 AM, John Danks wrote:

1. There are some textures that are removed like shinymetal. How to I
render the scenes using these materials? Will the material "metal" do?

Metal should work, but you'll have to look at the parameters for both the old material and the new one to get a good match.

Shinymetal has turned into a bit of a hack, now in this day where there's good spectral measured data from real metals.

The scenes/spds/metals/ directory has measured spectral data for a ton of different metals--if you can remember your chemical symbols, it should be easy to figure out (Ag for silver, etc.)  The metal.pbrt file shows how to use the new one.

dragon.pbrt
--------------

dragon.pbrt(81): Warning: Parameter "directphotons" not used
dragon.pbrt(81): Warning: Parameter "directwithphotons" not used

Shooting photons: [+++++++++++++++++++++++++++++++++         ]
(224.2s|61.3s)  terminate called after throwing an instance of
'std::bad_alloc'
 what():  std::bad_alloc
Aborted

Dragon is a _very_ large model. You are most likely running out of memory.

I think it's running out of memory, but I think that's due to a bug--I'll investigate and see if my hypothesis is right (but probably not until over the weekend...)  The new photon map integrator doesn't support doing direct lighting with photons, so those parameters are gone.  But it does store direct lighting photons during photon shooting, since they're needed as part of the computation that's done for final gathering.  Anyway, the important part is that it just stores as many direct photons as it finds and never stops (under the assumption that once enough other kinds of photons are found, that's a fine place to stop for direct.

The problem is that if it has to shoot a lot of photons to find caustic paths, say, it may store a ton of direct photons along the way, way more than one would want, purely because it had to start many paths to find a smaller number of caustic paths.  So my guess is that it's running out of memory due to that..  (Which I'd say is definitely a bug!)

Thanks,

-matt

[Matt Pharr]

On Oct 16, 2009, at 5:11 AM, Rachit Agrawal wrote:

dof-dragons.pbrt
---------------------

geometry/dragon.pbrt(1309069): Warning: Material "bluepaint" unknown.
geometry/dragon.pbrt(1309069): Error: Unable to create material
"bluepaint"

geometry/dragon.pbrt(1309069): Warning: Material "clay" unknown.
geometry/dragon.pbrt(1309069): Error: Unable to create material "clay"

The Lafortune BRDF stuff is gone from the system now (with time, it's become apparent that it's not a very good representation for measured BRDF data, and those materials were based on it..)  scenes/brdfs has some measured BRDF data from Cornell; there's some good paint data there at least.

tt-tire.pbrt
------------
geometry/tt.pbrt(885986): Warning: Parameter "Ks" not used
geometry/tt.pbrt(885986): Warning: Parameter "roughness" not used
geometry/tt.pbrt(885990): Error: Expected numeric parameter value for
parameter "roughness" with type "float".  Ignoring.
geometry/tt.pbrt(885990): Error: Expected numeric parameter value for
parameter "bumpmap" with type "float".  Ignoring.

Change those "float"s to "texture" to fix those errors.  That was a bug in the tt.pbrt file we distributed; pbrt v1 had a work-around that accepted it and printed a warning.  For v2 we're just reporting an error.  (And same for the Errors below....)

Thanks,

-matt