https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119108
--- Comment #7 from Matthew Malcomson <matmal01 at gcc dot gnu.org> --- FWIW I have managed to figure out what the difference between my internal build and the upstream one was -- my reproduction script has the line `-DCMAKE_BUILD_TYPE=Release` in it and the local build that I did some performance analysis on does not. >From looking at the build logs it seems the only real difference due to this difference in flags is that `-DNDEBUG` is passed to the compiler. So things still got optimised -- though obviously this is not the best for a benchmark run. However it does seem somewhat useful that without the abvoe `cmake` argument I can see the (now not inlined) `FindMatchLengthPlain` function change and start to take up a much greater proportion in the perf statistics with 68326d5d.
