Dynamical simulations of planetary systems can take weeks or even months to run. There is no way to efficiently parallelize N-body simulations with a small number of particles using GPUs or CPU clusters because they are inherently sequential and any communication overhead quickly dominates the runtime. In this talk, I will present the new symplectic N-body integrator WHFast512. WHFast512 uses Single Instruction Multiple Data (SIMD) parallelism and 512-bit Advanced Vector Extensions (AVX512) to speed up integrations of planetary systems by more than a factor of 10x, making it by far the fastest N-body integrator for planetary systems currently available. I will not only talk about N-body simulations but also give a short introduction on how to write programs using these AVX512 instructions so that you can optimize your own algorithms as well.
![](https://astronomy.yale.edu/sites/default/files/styles/large/public/event-images/8bea6f33-6fc6-49a5-a494-43edd754dc63.jpeg?itok=pZKypgf3)