I had obviously expected that, but the large number of reflective surfaces [1] greatly reduces the performance. The author explained in the blog post [2] that it is a path tracing (Monte Carlo simulation of multiple random rays incrementally averaged) and probably it is hard to limit the number of bounces small while maintaining the image quality. Indeed, the source code [3] says a hard limit of 1000 bounces, not big but not that small either.
I had obviously expected that, but the large number of reflective surfaces [1] greatly reduces the performance. The author explained in the blog post [2] that it is a path tracing (Monte Carlo simulation of multiple random rays incrementally averaged) and probably it is hard to limit the number of bounces small while maintaining the image quality. Indeed, the source code [3] says a hard limit of 1000 bounces, not big but not that small either.
[1] Something like http://is.gd/cFDlq5
[2] http://scanlime.org/2013/04/zen-photon-garden/
[3] https://github.com/scanlime/zenphoton/blob/bea23c1/hqz/src/z...