Lies... and then benchmarks...

[Ismael Venegas Castelló]

Continues: https://github.com/JuliaLang/julia/issues/14808#event-527206420

[Ismael Venegas Castelló]

@ylluminate looking at the structure of the Julia programs, I can only see @kostya is timing both Julia startup, the JIT and the code at the same time, which is wrong

• https://github.com/kostya/benchmarks/blob/master/base64/test.jl
• https://github.com/kostya/benchmarks/blob/master/base64/run.sh#L25-L26

Let's continue this discussion at julia-users:

• https://groups.google.com/forum/#!topic/julia-users/4AgBccftrps

El martes, 26 de enero de 2016, 16:36:22 (UTC-6), Ismael Venegas Castelló escribió:

Continues: https://github.com/JuliaLang/julia/issues/14808#event-527206420

[Ismael Venegas Castelló]

I was also saddened to see this observation:

Julia instead is slowing down and it disappeared from my radar during the last 9 month. It is still young and promising but it slowed down a bit this year so I don’t think it deserve the Top 5 for now.

http://www.davideaversa.it/2015/12/the-most-promising-languages-of-2016

@ylluminate: that is just one opinion! ;)

[Sisyphuss]

I didn't know that a man can learn so many languages... is he a genius? 

[Ismael Venegas Castelló]

I wondered the same first time I saw quine-relay! ...Yusuke is genius :D

* https://github.com/mame/quine-relay

[Sisyphuss]

See what I find :"yes, i am bad at julia, pr please how to better code in it. or i can remove the tests with julia."

Ref: https://github.com/kostya/benchmarks/issues/9

[Sisyphuss]

Didn't imagine that the Scheme is called "gauche" in Ubuntu...

P.S. "gauche" is a French word, which means "left"

[Ismael Venegas Castelló]

I'll try to fix one of them in my spare time, and see if he can fix the rest.

[Ismael Venegas Castelló]

That is just one implementation of scheme among many, try:

julia --lisp

;-)

[Sisyphuss]

Yeah, this is Jeff's Femato !

[Stefan Karpinski]

There are two issues here:

1. Literal speed – the benchmarks indicate Julia being 2x slower than fast languages. They benchmarks, however, are including Julia startup time and JIT compilation. So being 2x slower than C is pretty damned good.

2. Development speed – "Julia instead is slowing down" is not, unless I'm seriously misreading it, about performance regressions, but about development velocity and rate of adoption. There is certainly less hype around Julia right now than a year or two ago, judging e.g. by number of Julia-related posts on HN per unit of time. On the other hand, there are far more serious users of the language (the NY Fed, for one obvious example), and the growth of the ecosystem in terms of number of packages continues in the same linear fashion as it has since we introduced a package manager.

In terms of development of the language itself, I can see how there could be an impression that it's slowed down since the features being tackled these days are much larger and more difficult. As a result, the time between them is correspondingly longer. But 0.4 introduced function call overloading, docstrings, generated functions – a feature that no other mainstream language seems to have, which is nearly a superpower for writing general and efficient numerical libraries – package precompilation, and a new, much faster garbage collector, just to name a few.

The 0.5 release is going to introduce many even more impressive features, including: a number of important changes to the behaviors of arrays and strings, a revolutionary improvement in the performance of higher order programming – you'll have to try it to really understand what a big deal this is – real multithreading as an experimental feature, and much better support for static whole-program compilation. These are huge features that don't exist in any other mainstream dynamic language and it's vital to tackle them in Julia while the language is still malleable enough that we can. Think about how many early decisions in other languages ended up making features like multithreading and static compilation impossible. In a year, radically redesigning the way closures and higher-order functions work would be far more difficult, and maybe impossible.

Perhaps Julia is in a lull in the hype curve at the moment. But if so, I think that's ok – we're doing crucial work in preparation for Julia 1.0 – work that will let us race far ahead into the future.

[cdm]

epic.