it seems it's a bit 50-50 in terms of performance gains, but the main gain, from what i understand, is the ability for the racket developers to code in racket and chez scheme now instead of c.
Note that this report is 9 months old. I guess there will be an new report this January. There has been a lot of small improvements.
I'm not tracking all of them, but I think that now there are less checks to ensure that the functions return don't return multiple values in the wrong moment, and there are less unnecessary continuation marks.
As Racket CS slowly improves towards matching the mainline implementation, those results can give the impression that Racket CS is slow. To put things in context, look where Racket was already at in comparison to another very dynamic language, Python:
I don't know how to directly compare Python 3 and Racket CS on that site, but comparing three ways paints Racket CS in a more favorable light. Hopefully Racket CS will close the gap with Racket, but it’s already doing pretty great relatively... unless we compare to SBCL, Rust, Haskell, etc.
Comments
Does this mean Racket CS is faster than the default Racket?
sometimes yes, sometimes no.
https://blog.racket-lang.org/2019/01/racket-on-chez-status.h...
it seems it's a bit 50-50 in terms of performance gains, but the main gain, from what i understand, is the ability for the racket developers to code in racket and chez scheme now instead of c.
Note that this report is 9 months old. I guess there will be an new report this January. There has been a lot of small improvements.
I'm not tracking all of them, but I think that now there are less checks to ensure that the functions return don't return multiple values in the wrong moment, and there are less unnecessary continuation marks.
It does not mean that. It just means Racket CS is faster in 7.5 than it was in 7.4.
fwiw
https://benchmarksgame-team.pages.debian.net/benchmarksgame/...
As Racket CS slowly improves towards matching the mainline implementation, those results can give the impression that Racket CS is slow. To put things in context, look where Racket was already at in comparison to another very dynamic language, Python:
https://benchmarksgame-team.pages.debian.net/benchmarksgame/...
I don't know how to directly compare Python 3 and Racket CS on that site, but comparing three ways paints Racket CS in a more favorable light. Hopefully Racket CS will close the gap with Racket, but it’s already doing pretty great relatively... unless we compare to SBCL, Rust, Haskell, etc.
Note that several of these benchmarks use 4 cores on Python but only one on Racket. Time to revisit the Racket programs and use places.
Look at the single core benchmarks to get a more 1-1 comparison.
Please contribute improved programs:
https://salsa.debian.org/benchmarksgame-team/benchmarksgame/...
Please note the URL I provided shows a comparison between Racket and Racket CS (not Python).