Hacker News Reader

Comment by saidajigumi | Hacker News Reader

So if I understand correctly, the pidigits test in this case this isn't a self-contained, fully ported microbenchmark, but rather a test of some random Rust code vs. an optimized library[1]?

If that's true, then what's the point of including pidigits if the entire algorithm isn't going to be ported into all target languages?

[1] GMP, The GNU Multiple Precision Arithemtic Library: https://gmplib.org/

View on HN

As I understand it, the rule is "if your language comes with a bignum implementation, you must use it, otherwise you may use GMP". I've been seriously tempted to suggest removing our bignum implementation for this reason (though I won't).

Python uses GMP while having a bignum implementation. I believe any language is allowed to use it.

Python: http://benchmarksgame.alioth.debian.org/u64/program.php?test...

Ruby: http://benchmarksgame.alioth.debian.org/u64/benchmark.php?te...

There is no such rule.

When someone contributes a Rust pi-digits program that uses GMP it will be shown (as long as it uses the same algorithm).

We get to see both programs that use GMP and programs that don't.

It's more of a "testing the platform" rather than the language. You can write a Rust program that runs at exactly the same speed as a C program (the difference here would be clang backend vs. gcc moreso than the choice of language).

That becomes more of a pass/fail test (either you're using the fastest known library or you aren't) which isn't very interesting as a benchmark.

Reply on news.ycombinator.com