One of the problems plaguing the distributed version of Illumina PRT is that of resource management. Precisely, when multiple task groups are running concurrently, switching resources from one task group to another can be a great time sink. A PE’s current data set (geometry, acceleration … Continue reading Particle file format
These last weeks I was working on optimising the Interactive Global Illumination (IGI) integrator for Illumina. The implementation uses a technique known as interleaved-sampling, where the image space is divided into square regions, usually of 3×3 or 4×4 pixels, and the rays shot from the viewer through … Continue reading An update long due
In the last few days I’ve been working on integrating image-based lighting into the Illumina renderer. At the moment, I have a preliminary setup running on light probes from low-dynamic range (LDR) images. The light probes are sampled using a simple cosine-weighted distribution which is far from optimal. In fact, when low sampling rates are used, output suffers from excessive variance. Next step is that of implementing a better sampling strategy…
Panagia Angeloktisti Church (Kiti) model courtesy of Jassim Happa, University of Warwick.
In the last couple of months I’ve been doing some work on a cross-platform distributed renderer, Illumina PRT. The renderer is still in its infancy and requires a substantial amount of work before it can be called decent by any standard. Hopefully I’ll be documenting … Continue reading Illumina PRT – A distributed physically-based renderer
For the last few days we’ve all been working on wrapping up the projects and writing the dissertations which are due in three weeks’ time. I have created an application object for Vistas, allowing users to create and set up applications easily and quickly. Consequently I built a basic sandbox on top of the application object which takes the onus of additional setup and world manipulation off the user’s shoulders.
Colin has been building his own Gravitas sandbox (based on the Vistas Sandbox), which additionally allows one to load physics scenes from a file based on a custom text format. The loader creates alternative representations of the physics counterparts in Vistas using the MeshForge object.
The video below shows the first fruits of the simulation platform:
And following are some screen shots…
More than two months have gone by since I’ve last updated the Vistas progress log, and during this time, besides sitting for my exams (and passing all of them, thank God), I’ve started closing the first iteration of this project. I am pretty happy with the outcome, and more so considering the constraints within which research, design and development took place.
Since the last update, I’ve created a runtime typing system for better scene management and extension, modified the high level rendering pipeline, and re-written the effect system for the umpteen time. Although the former effect system was good, it was way too complex. The new version is simpler, but no less powerful.
I have also added an application framework for effortlessly setting up Vistas-based applications; now I’m working with Colin and Gordon to prepare some sort of sandbox within which we can have a unified graphics-physics-scripting system we can meddle with quickly and easily. The reason we’re working on a sandbox at this stage is simply that we need a showcase that is hassle free to set up.
Anyway, I’ll try to post in somewhat more detail next time. Before I leave I think credit is due to Mr. Paul Bourke for the cube map used in the screen shots above. As to what regards the Audi model, I don’t know who the author is – Ben, my brother, sent it to me, but he said he didn’t do it… oh, and thanks to Colin for UV mapping the model; it would have taken me ages to do.