GCC 10 Link-Time Optimization Benchmarks On AMD Threadripper

Written by Michael Larabel in Software on 7 January 2020 at 03:08 PM EST. Page 2 of 4. 27 Comments.

With HPC Challenge there were some minor improvements in performance particularly when pairing LTO with the "-fwhole-program" option.

For FFTW the results were mixed with LTO leading to lower performance with the smaller FFT sizes but improving with the larger ones.

The MrBayes Bayesian inference in phylogeny software package was benefiting from LTO on GCC 10.

The BYTE Dhrystone 2 benchmark benefited a great deal from Link-Time Optimizations, coming out about 40% faster.