I am using approximately 1,000,000 photons for each of these renders. Photons are sampled from a 1-unit sphere. Also, based on the shape of my light and so forth, it's evident that some of my random sampling is biased.
Image without any path tracing (Just over a minute before optimizations; ~25 seconds after optimizations)
Image with one bounce of path tracing (About 17 minutes; took ~45 minutes before optimizations)
Blank
Some specs
Different specs
Finally actual good-ish looking photons
Now it has roughly enough photons
Higher resolution (and darker) render. Still no photon bouncing.
Photon size is larger
Now photons are bouncing multiple times (partially incorrectly)
Higher-resolution render
Tweaking more parameters
Smoother
Trying more photons
Added teapot
Trying a layer of path tracing (got it backwards)
Got path tracing right (but not really) but forgot inverse-square falloff for direct illumination
Path tracing is still too dark for GI (Photon map too dark)
Trying scaling photon map by 4 * pi (area of sphere)
Pure GI (no direct)
Adding back in direct illumination
Photon map for 2nd bounce of path tracing
Pure photon mapping
Pure photon mapping
Less-blotchy photon map
Photon map render for debug
Progress on photon precomputed irradiance
(More)
(More)
Now it looks good again. Also, the light distribution on the ceiling is a circle, not a square (but I find out later it's still not correct)
But now there's splotches!
And now there's a crack in the wall!
(More)
(More)
(More)
(More)
Cracks in the wall are gone
Cracks are fixed, but things are still splotchy
Doing a render with path tracing
Another render (too many bright spots!)
Trying diffuse/specular only (no reflections or refractions)
Trying with only 1 bounce
Problems resolved; now spherical sampling is actually uniform
MacBook Pro (15-inch, 2016)
2.7 GHz quad-core Intel Core i7 Skylake (6820HQ)
16 GB 2133 MHz LPDDR3 RAM