(Replying to PARENT post)

I find it remarkable that one person with just over a year of Rust experience can successfully rewrite ~6000 lines of well-tested, highly tuned Fortran in Rust and end up with similar performance on benchmarks without having in-depth expertise in the domain (simulation of fluid dynamics). The benchmarks (https://mckeogh.tech/post/shallow-water/84896735-54a97580-b0...) show similar performance across all input sizes for both the old Fortran and the new Rust code, with Fortran outperforming in some cases and Rust outperforming in others.

👤cs702🕑5y🔼0🗨️0

(Replying to PARENT post)

Actually closer to 4 years of experience:

> As for my Rust experience, I’ve been using it for personal projects since ~2016 and I worked as a Rust software engineer at a startup in Berlin for a year after leaving high school.

Publish date on this blog post: July 14th, 2020

👤faitswulff🕑5y🔼0🗨️0

(Replying to PARENT post)

FTA: “The result of this is that these simple operations are being performed on some very large amounts of data which I believe is resulting in a memory bandwidth bottleneck. This hypothesis explains why SIMD did not improve performance and performance is not significantly improved by parallelisation”

If that’s true, comparing system load or power usage could be a way to differentiate between the two implementations.

👤Someone🕑5y🔼0🗨️0

(Replying to PARENT post)

Rust is not a very good example here considering the amount of time the op might have spent waiting for the compiler and steep learning curve of the language...

👤tastyminerals🕑5y🔼0🗨️0