anthony-fiddes / raytracing-1w Goto Github PK

This is a simple ray tracer I built following "Ray Tracing in One Weekend".

Here is an image I generated using the code snippet that the authors used to produce their book cover:

And here are some benchmarks showing the difference in performance between the parallel and non-parallel renders:

goos: darwin
goarch: arm64
pkg: github.com/Anthony-Fiddes/raytracing-1w
BenchmarkRenderSimple-10                   	     30	159664206 ns/op	   2106 B/op	     14 allocs/op
BenchmarkRenderSimpleParallel-10           	     30	104225461 ns/op	  13358 B/op	     42 allocs/op
BenchmarkRenderRandomSpheres-10            	     30	4738820744 ns/op  60009 B/op	    960 allocs/op
BenchmarkRenderRandomSpheresParallel-10    	     30	808697338 ns/op	  69781 B/op	    985 allocs/op
PASS
ok  	github.com/Anthony-Fiddes/raytracing-1w	180.542s

So that's ~1.5x faster for the simple scene and ~5.86x faster for a more complicated scene that looks like the render above. These benchmarks are basically shrunken down versions of renders from the book so they don't take forever to complete a run.

I ran this on an M1 Max with 8 performance cores and 2 efficiency cores, so the peak theoretical increase would have been somewhere around 8x. Realistically the performance increase is even less than reported above since there's likely to be throttling if a render goes long enough.

Another note: while writing the benchmarks, I realized that my implementation will degrade in performance when the number of samples is not greater than the number of cores on the system used, but that shouldn't be much of an issue since using a low number of samples gives a very grainy image. In any case, if I were to revisit the project it could be interesting to consider a different concurrency plan or investigate making use of GPU acceleration.