@RazielQuote:
This version was compiled with maximum optimizations (-O3), Link Time Optimization (-flto -flto=jobserver) and dead code elimination (-ffunction-sections -fdata-sections) and as a release build, so as fast as it will currently get.
I wonder, have you ever checked / profiled such settings? In my experience, ScummVM/m68k and gcc 13.2 didn't react very well to -ffunction-sections -fdata-sections (the resulting binary was *larger* and no, I did not forget -Wl,--gc-sections) and when reading about the effect of LTO, it also often results in more bloated code.
And then of course -O2 vs. -O3... that often leads to surprising (in a negative sense) results.
So for my ScummVM/m68k port, I stayed faithful to the classic -O2 -fomit-frame-pointer and a few classics (-ffast-math, -fno-exceptions, NDEBUG etc).