Small unrelated nitpick: Instead of `epochTime` for benchmarking you can use
`times.cpuTime` or `monotimes`
@Araq \- thanks for the precision - I'll focus on using gc:arc on latency and
memory-sensitive applications.
I consider it production ready as a replacement for `--gc:none` and embedded
development. It's known to still be slower than the default GC in many cases
(as you've seen it here) but more optimizations are coming. You test for
throughput here though, arc focusses on latency and memory consumptio
Hi
is \--gc:arc considered ready for production use in nim 1.2? I notice
considerably worse performance of some code compiled using \--gc:arc. Here's an
example:
import times, strutils
type
OpType = enum Inc, Move, Loop, Print
Ops = seq[Op]
Op = object