https://ecraven.github.io/r7rs-benchmarks/
MIT Scheme 9.2.1 is fast.
I'd like to see where Guile now with version 3. It would be nice to have Stalin actually compile, too.
>>2
Stalin is gone for good. Not only is it total abandonware but it only supports a subset of R4RS. I have no idea why it's included in R7RS benchmarks.
>>3
On the benchmarks published last year, Stalin was still finishing 38/57 tests (that's 3 more than Chicken interpreted) and acing a lot of benchmarks. In some tests, Stalin was like 10 times faster than the 2nd Scheme (Array1, fft, mbrot). https://web.archive.org/web/20190212174828/https://ecraven.github.io/r7rs-benchmarks/
Sure it's R4RS but it would be hard to optimize it further.
How did Stalin go from finishing 38 benchmarks in 2018 to not finishing any in 2019- they are the same benchmarks, right?
>>5
I don't know, but if I remember they were still R6RS benchmarks.
Racket did a great leap forward with the switch to Chez Scheme.
How do I make my favourite implementation win the benchmarks? Do I need to become a compiler engineer?
Updated on 2022-06-19: https://ecraven.github.io/r7rs-benchmarks/
How Cisco Sirs could achieve Chez being so fast?
It's a professional product, unlike the other implementations that are just toys for hobbyists.
I prefer Guile, is it pro?
nope
I don't think anyone is paid to work on Guile but I am not sure.
Aren't GNU developers paid? I don't think they are just volunteers.
>>15
There are paid managers who are project leads and they are probably the most active contributor to their projects.