22-perf

Using the Go execution tracer

Watch the episode here:

justforfunc 22: using the Go tracer

Or read the blog post here.

What is the fastest way to compute a Mandelbrot set like the one below?

The final times are

mode	seconds
sequential	4.669
pixels	3.004
cols	0.689
workers	2.967
workers + buffer	1.226
col workers	0.714
col workers + buffer	0.698

You can run the benchmarks yourself by running:

$ go test -bench=.

If you want to run only the benchmarks for given size (e.g. 64) use:

$ go test -bench=./64
goos: darwin
goarch: amd64
pkg: github.com/campoy/justforfunc/22-perf
BenchmarkSeq/64-8                            300           4534800 ns/op
BenchmarkPixel/64-8                         2000           1083436 ns/op
BenchmarkCol/64-8                           2000            905426 ns/op
BenchmarkWorkers/64-8                        500           2698059 ns/op
BenchmarkWorkersBuffered/64-8               2000           1114785 ns/op
BenchmarkRowWorkers/64-8                    2000            772574 ns/op
BenchmarkRowWorkersBuffered/64-8            2000            745455 ns/op
PASS
ok      github.com/campoy/justforfunc/22-perf   13.190s

Also, modify the constant named complexity in the pixel function to compute simpler or more complex fractals.