This is the source code for the renderer implementation described in the paper FlexRender: A Distributed Rendering Architecture for Ray Tracing Huge Scenes on Commodity Hardware, presented at GRAPP 2013. FlexRender distributes geometry among a cluster of commodity machines to keep the scene in-core within the cluster, and passes rays around as messages using a feed-forward messaging architecture that never requires replies. This keeps communication and rendering completely asychronous.

For more information about FlexRender, including the paper and slides from the presentation, see flexrender.org.

We rely on several third-party libraries, but anything that you won't find on a standard Linux distro is downloaded, built, and statically linked by our dependency build script. To build FlexRender itself, you'll need a C++ compiler with reasonable C++11 support (GCC 4.7+ is known to work) and Premake, a build script generator.

First, check out the source code and run the dependency build script. This only needs to be done once. (No packages are installed, everything is built and linked within the 3p/ directory so no root privileges are required.)

git clone https://github.com/bobsomers/flexrender.git
cd flexrender/3p
./make.bash

Next, generate GNU Makefiles for building FlexRender by running Premake. Then, run make (optionally with config=release for a release build).

cd ..
premake4 gmake
make config=release

This should leave you with 3 binaries in the bin/ directory. The flexrender and flexworker executables are the renderer and worker respectively. The baseline is the image plane decomposition.

• 3p/ All third-party libraries and build scripts.
• frlib/ Lua libraries for scene files and FlexRender shaders.
• scenes/ Some example scenes and shaders.
• scripts/ Handy scripts for profiling.
• src/[baseline|render|worker] Code specific to the baseline, renderer, and worker executables.
• src/shared Shared code for the libfr static library.
• config.lua Example renderer configuration.

In two shells, we run two flexworkers. One on the default port (19400), the other on port 19401.

bin/flexworker
bin/flexworker -p 19401

In a third shell, we run the renderer, giving it the path to its renderer configuration (config.lua) and the scene file.

bin/flexrender config.lua scenes/cornell.lua

When finished, this should write out cornell.exr. It will also write out the image buffers of each worker, as well as a CSV file for each worker with rendering statistics.

The FlexRender source code is licensed under the MIT license. We encourage you to use the code, ideas, and techniques to enhance your own research.

If anything from FlexRender does aid your research, please cite the paper appropriately.