2 ms·
I don't think Julia was designed for pure overhead projects in memory-constrained environments, or for squeezing out that last 2% of hardware performance to cut
by wrathofmonads 1y ago
I don't think Julia was designed for pure overhead projects in memory-constrained environments, or for squeezing out that last 2% of hardware performance to cut costs, like C++, Rust or Zig.
Julia is the language to use in 2025 if what you’re looking for is a JIT-compiled, multiple-dispatch language that lets you write high-performance technical computing code to run on a cluster or on your laptop for quick experimentation, while also being metaprogrammable and highly interactive, whether for modelling, simulation, optimisation, image processing etc.
- throwawaymaths 1y agoactually I think it sort of was, I remember berkeley squeezing a ton of perf out of their cray for a crazy task because it was easy to specialize some wild semi-sparse matrix computations onto an architecture with strange memory/cache bottlenecks, while being guaranteed that the results are still okay.
- leephillips 1y agoYes. Remember that the four languages that have been used for very high performance computing are Fortran, C, C++, and Julia: https://www.hpcwire.com/off-the-wire/julia-joins-petaflop-club/ https://www.hpcwire.com/off-the-wire/julia-joins-petaflop-cl... For large-scale physics simulations, Fortran or Julia are the obvious choices.