4 ms·
actually 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
by throwawaymaths 1y ago
actually 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.