Benchmark produce

[yebai]

No description provided.

[rikhuijzer]

I've added a few suggestions to make the code more idiomatic. On top of that, the code contains a lot of semicolons at line endings. Would be good to remove those too

[yebai]

I've added a few suggestions to make the code more idiomatic. On top of that, the code contains a lot of semicolons at line endings. Would be good to remove those too

thanks @rikhuijzer - this PR contains a lot of code copy-and-pasted from terminals when I was debugging #135, which is a bit messy.

[rikhuijzer]

In general, I like this PR by the way a lot. I had a lot of trouble debugging the functionality when I was working on Libtask. One question: do you know how well these estimates represent the workload under normal load?

[yebai]

One question: do you know how well these estimates represent the workload under normal load?

These examples cover loops and produce calls, which I suspect are the bottlenecks for optimal performance. The former is very sensitive to type-instability, while the latter tests the efficiency of the most frequent TapedTask operation. The TArray stuff is potentially a bottleneck for performance, but they can be depreciated now since the new tape-based mechanism provides a more flexible solution for controlling which variables get copied.