News: 0000827942

Version 1.5 of the [1]Julia programming language has been [2]released . On the Julia blog, Jeff Bezanson and Stefan Karpinski [3]describe the highlights of the release , which includes struct layout improvements for decreasing heap allocations, stabilization of the multithreading API, faster random numbers, changes to the scoping rules in the read-eval-print loop (REPL), and more. " Julia excels at simulations, so random numbers are important to a lot of users of the language. For this release [4]Rafael Fourquet , one of the primary architects of the Random standard library and a prolific contributor in general, implemented some impressive algorithmic improvements for some popular cases. The first is a major improvement when generating normally-distributed double-precision floats. Calling randn(1000) is nearly twice as fast in Julia 1.5 compared with Julia 1.4. Generating random booleans also got much faster: rand(Bool, 1000) is nearly 6x faster. Finally, sampling from discrete collections has also gotten faster: rand(1:100, 1000) got 25% faster. " LWN looked at Julia ( [5]part 1 , [6]part 2 ) back in 2018, shortly after the release of Julia 1.0.



[1] https://julialang.org/

[2] https://discourse.julialang.org/t/julia-v1-5-0-has-been-released/44169

[3] https://julialang.org/blog/2020/08/julia-1.5-highlights/

[4] https://github.com/rfourquet

[5] https://lwn.net/Articles/763626/

[6] https://lwn.net/Articles/764001/