GCC 10 vs. GCC 11 Compiler Performance On AMD Zen 3
With FFTW the runs with "-march=native" at both optimization levels tested were yielding slightly better performance with GCC 11.0. The link-time optimized run was obviously the fastest but its performance was flat or even slightly lower with the FFTW benchmark on GCC 11.
In the case of the MrBaues performance, with the highest optimizations the performance on GCC 11 was slightly behind that of GCC 10.
Other workloads like Zstd were also pointing to GCC 11 being potentially a tiny bit slower than GCC 10, at least with the testing on this AMD Ryzen 9 (Zen 3) desktop.
The SciMark 2.0 results were looking positive for GCC 11 with those tests being prone to highlighting compiler differences.
Or in the case of the TSCP chess benchmark it was showing nice GCC 11 performance improvements across the board.