I tested with tcc, cparser and my own slimcc, the benchmark showed multiple process with htop, so there are some dynamic cpu scaling randomness going on. I saw

I tested with tcc, cparser and my own slimcc, the benchmark showed multiple process with htop, so there are some dynamic cpu scaling randomness going on. I saw larger run-to-run differences with the same binary, than with comparing cherry-picked best-run of each, so probably not the best benchmark.