top | item 30245275

(no title)

isofruit | 4 years ago

How can such a benchmark be in any way informative? How much am I allowed to optimize the nim code? How much am I allowed to optimize the Rust code? Am I allowed to just change nim's GC to ARC/ORC? I'm sorry but this just does not make any sense to me.

discuss

onebot|4 years ago

It is more informative than nothing. It shows different languages implementing the same functions and the performance of those. It is something to compare to. Maybe they can all be optimized, etc. But you need to have some kind of starting point. I think people tend to poo-poo benchmarks (unless their fav language at top) and claim they don't represent real world. But I always argue that they at the very minimum give you some context for comparison. If you look at how they are implemented you can get further context. I only wish the benchmarks had some metric for how idiomatic the implementation is compared to normal developers.

igouy|4 years ago

Yes.

https://benchmarksgame-team.pages.debian.net/benchmarksgame/...

----

> I only wish the benchmarks had some metric for how idiomatic the implementation is compared to normal developers.

Maybe the benchmarks game website kind-of does — source code size.

Otherwise "I only wish…" so — wishful thinking.

Otherwise "how idiomatic" seems to mean — how I would implement the program today, if I didn't care [goldilocks] too much about performance?

(Incidentally "normal developers" doesn't seem to be stretching towards inclusive.)

gameswithgo|4 years ago

All benchmarks are informative, also all benchmarks are imperfect and incomplete, for reasons you have specified and many others.

However it is a mistake to then suppose that nothing can be known and just go ahead and use Python to make a AAA first person shooter.

Mikeb85|4 years ago

On the flipside slow languages (Python included) are used to script some AAA games.

And no language other than C++ has really found traction for stuff like engines; not even Rust...