Make Cycles better for render research

All papers that I read uses pbrt or mitsuba as a base for its research. Many of this researches releases code in GPL 3 licence. What if cycles could be somewhat easy to be used in this researches. Maybe a 'small-cycles'. Anything that facilitates later incorporation of the new algorithm in cycles.

My suggestion is create a 'small-cycles', with only basic features. In pair with mitsuba and pbrt in terms of light and materials. This 'small-cycles' should form the code base of the complete cycles. Maybe should be necessary implement also some algorithms like Photon Map and/or VCM, as the main purpose of the research is 'how fast my new idea converges to a perfect solution of the 'rendering equation''. This other algorithms may be or not part of the 'complete cycles'. I think initially no.

The second 'phase' of the project should be a campaign 'Do your research in small-cycles. Your good idea in production, fast!'. Send letters to universities, call researches in Epic, Intel, Nvidia, etc. This maybe a special paper for Ton Rosendal.

If this project is successful, the best and latest algorithms for rendering will be soon available to blender community, without the cost to re-implement them. Only latter, maybe, a optimization phase can be made by cycles developers, but most of code will be already there.