The Computer Language Benchmarks Game

[qznc via Digitalmars-d]

On Sunday, 7 August 2016 at 17:53:12 UTC, Isaac Gouy wrote:
> On Sunday, 7 August 2016 at 05:04:29 UTC, Walter Bright wrote:
>
>>> Ultimately, my opinion is that the benchmark is outdated and 
>>> not useful today. I
>>> ignore it, if anybody cites the benchmark game for 
>>> performance measurements.
>>
>> Yeah, I wouldn't bother with it, either.
>
>
>
> Rather than only being dismissive, wouldn't it be more 
> effective to say what you would bother with.
>
> https://github.com/kostya/crystal-benchmarks-game
>
> https://github.com/def-/nim-benchmarksgame
>
> Where are the D comparison programs that people should bother 
> with?

Comparing so many languages is too broad. We need more specific 
benchmarks which must be looked at in more detail.

For example, it makes sense to have a benchmark for regular 
expressions. It is interesting to compare Rust and D, because 
they both use compile-time tricks, which C cannot do. It makes no 
sense to compare them to an equivalent Python/Ruby/etc program, 
because they all use the same C library internally. More details: 
Regular expressions are used in very different scenarios, so one 
program might not be enough. Simple expression or complex 
expression? Find first or find all? ASCII or Unicode? Large 
haystack or small?

If you want to answer the big question of "how fast is D compared 
to C", the best answer is "roughly the same if you compare clang 
to LDC, because they use the same optimizer and code generator". 
Getting some numbers from a bunch of benchmark programs does not 
improve the answer. You can only get a better answer in the sense 
that it becomes more specific to your specific use case, but that 
requires you to build your own benchmark.