'Ignis' is a high-performance raytracer implemented using the Artic frontend of the AnyDSL compiler framework (https://anydsl.github.io/) and based on Rodent (https://github.com/AnyDSL/rodent). The renderer is usable on all three major platforms (Linux, Windows, MacOs).
Some scenes rendered with Ignis. Acquired from https://benedikt-bitterli.me/resources/ and converted from Mitsuba to our own format. Both images took roughly one minute to render. With an RTX 2080 Super you can even have an interactive view of the scene.
A sample scene from https://github.com/KhronosGroup/glTF-Sample-Models directly rendered within igview
.
- AnyDSL https://github.com/AnyDSL/anydsl
- Eigen3 http://eigen.tuxfamily.org
- Intel® Threading Building Blocks https://www.threadingbuildingblocks.org/
- ZLib https://zlib.net/
The following dependencies will be downloaded and compiled automatically. Have a look at CPM for more information.
- Catch2 https://github.com/catchorg/Catch2
- imgui https://github.com/ocornut/imgui
- imgui-markdown https://github.com/juliettef/imgui_markdown
- PExpr https://github.com/PearCoding/PExpr
- pugixml https://github.com/zeux/pugixml
- RapidJSON https://rapidjson.org/
- Simple Tag Preprocessor https://github.com/PearCoding/stpp
- stb https://github.com/nothings/stb
- tinyexr https://github.com/syoyo/tinyexr
- tinyobjloader https://github.com/tinyobjloader/tinyobjloader
- tinyparser-mitsuba https://github.com/PearCoding/TinyParser-Mitsuba
Ignis is available on docker hub with some preconfigured backends! pearcoding/ignis
More information is available here docker/README.md
If you made sure the required dependencies are installed in your system, create a directory to build the application in:
mkdir build
cd build
Next step would be to configure and finally build the framework. You might use your favorite generator (e.g. Ninja
)
cmake -G Ninja ..
cmake --build .
If Ninja
is not available, skip the -G Ninja
parameter. You can also set FETCHCONTENT_UPDATES_DISCONNECTED
to ON
to speed up the cmake steps after the initial cmake configuration.
More information about building Ignis on other systems are available at the documentation online or in the offline version of the documentation in docs/
The raytracer has multiple backends available to make sure the best optimized kernel is used for certain tasks. Therefore, multiple device and feature specific modules, so-called drivers, have to be compiled.
The compilation process might take a while depending on your hardware and feature selection. Parallel compilation of the drivers is disabled by default. Multithreading might freeze your operating system due to the high memory and cpu use. You can use the CMake option IG_BUILD_DRIVER_PARALLEL
to enable it if you are sure your system can handle it.
The frontends of the raytracer communicate with the user and one, optimal selected, backend. Currently, four frontends are available:
igview
This is the standard UI interface which displays the scene getting progressively rendered. This frontend is very good to get a first impression of the rendered scene and fly around to pick the one best camera position. Keep in mind that some power of your underlying hardware is used to render the UI and the tonemapping algorithms. Switching to the UI-less frontendigcli
might be a good idea if no preview is necessary. Note,igview
will be only available if the UI feature is enabled and SDL2 is available on your system. Disable this frontend by setting the CMake optionIG_WITH_VIEWER
to Off.igcli
The commandline only frontend is the same asigview
but without any UI specific features and no interactive controls. In contrary toigview
,igcli
requires a maximum iteration or time budget to be specified by the user. Progressive rendering is not that useful without a preview. (We might add progressive rendering back, but I need a convincing argument for that...)igtrace
This commandline only frontend ignores camera specific information and expects a list of rays from the user. It returns the contribution back to the user for each ray initially specified.Python API
This simple python API allows to communicate with the runtime and allows you to work with the raytracer in interactive notebooks and more. The API is only available if Python3 was found in the system. You might disable the API by setting the CMake optionIG_WITH_PYTHON_API
to Off.
Use the --help
argument on each of the executables to get information of possible arguments for each frontend.
Each frontend requires the exact location of the backends to run. An automatic search procedure will detect them in the system. In some rare cases the automatic search procedure might fail to find all the backends. In that case the environment variable IG_DRIVER_PATH
can be used to point to the directories containing driver modules. The environment variable is similar to the PATH
variable used in Linux environments and should contain absolute paths only, separated by ':' if multiple paths are provided. Setting IG_DRIVER_SKIP_SYSTEM_PATH
will prevent the automatic search and only depend on IG_DRIVER_PATH
.
Run a frontend of your choice like this:
igview scene/diamond_scene.json
All available components are documented in the docs/
folder. A documentation can be created with
cmake --build . -t ig_documentation
from the build/
folder.
A quite recent version of the above documentation is available at: https://pearcoding.github.io/Ignis/
Ignis uses a JSON based flat scene description with instancing. Support for shading nodes is available via PExpr, image and procedural textures. A schema is available at docs/refs/ignis.schema.json
You might use the mts2ig
to convert a Mitsuba scene description to our own format. Keep in mind that this feature is very experimental and not all BSDFs work out of the box.
You can also use rad2json
to convert geometry used in the Radiance framework to our tool. Keep in mind that no BSDF and lights are mapped as the two raytracers are vastile different in these regards.
Ignis is able to understand glTF files. You can embed glTF files in Ignis's own scene description file or directly use the glTF file as an input to the multiple frontends.
A Blender plugin is planned for the future.
Two tiny tools exr2hdr
and hdr2exr
are available to convert between the Radiance favorite image format HDR to the advanced OpenEXR format and vice versa.
This is useful to ease the transfer from Radiance to our raytracer, but you can disable them by setting the CMake option IG_WITH_TOOLS
to Off.
Actually, the tool might convert from any format the stb_image framework supports to the second format...
The Ignis client has an optional UI and multiple ways to interact with the scene:
1..9
number keys to switch between views.1..9
andStrg/Ctrl
to save the current view on that slot.F1
to toggle the help window.F2
to toggle the UI.F3
to toggle the interaction lock. If enabled, no view changing interaction is possible.F11
to save a snapshot of the current rendering. HDR information will be preserved. Use withStrg/Ctrl
to make a LDR screenshot of the current render including UI and tonemapping. The image will be saved in the current working directory.R
to reset to initial view.P
to pause current rendering. Also implies an interaction lock.T
to toggle automatic tonemapping.G
to reset tonemapping properties. Only works if automatic tonemapping is disabled.F
to increase (or withShift
to decrease) tonemapping exposure. Step size can be decreased withStrg/Ctrl
. Only works if automatic tonemapping is disabled.V
to increase (or withShift
to decrease) tonemapping offset. Step size can be decreased withStrg/Ctrl
. Only works if automatic tonemapping is disabled.WASD
or arrow keys to travel through the scene.Q/E
to roll the camera around the viewing direction.PageUp/PageDown
to pan the camera up and down.Notepad +/-
to change the travel speed.Numpad 1
to switch to front view.Numpad 3
to switch to side view.Numpad 7
to switch to top view.Numpad 9
to look behind you.Numpad 2468
to rotate the camera.- Use with
Strg/Ctrl
to rotate the camera around the center of the scene. Use withAlt
to enable first person camera behaviour.
The project is funded by the Velux Stiftung and developed in cooperation with the Computer Graphics chair of the Saarland University, Fraunhofer Institute for Solar Energy Systems ISE and DFKI.