[eonil]

I don't think ray-tracer is a good example for Go.

Ray-tracer workload can be fully split into each hardware core. Only input data needs to be shared, and even the input mostly doesn't make any race condition. The algorithm even can run on GPU.

This is handicap for Go. Go wants to solve - safe and easy concurrency without race condition for complex logic. So (IMO,) Go needs to make some overhead (or sacrifice some performance feature) for its goal. But in ray-tracer example, this ability mostly not required.