In the course unit Energy Efficient Computing, ZITI students Maximilian Matthias Mielke and Leandro Rischard Borzyk achieved an impressive 1300x speedup for the problem of matrix-matrix-multiplication on a Raspberry Pi over a basic implementation that was just relying on compiler optimizations only. This was possible by various code transformations, memory prefetching, using SIMD, assembly, multiple cores, and finetuning with the help performance counters.
