https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79149
--- Comment #5 from Arnd Bergmann <arnd at linaro dot org> --- -fno-schedule-insns is comparable in stack frame size to "-fno-sched-critical-path-heuristic -fno-sched-dep-count-heuristic" on all architectures (give or take a few bytes), but actually produces much better code. In my simulated mips64 run, I see these numbers: -O2: 49.0Mbit/s -O2 -fno-sched-critical-path-heuristic -fno-sched-dep-count-heuristic: 109.7 Mbit/s -O2 -fno-schedule-insns: 179.2 Mbit/s The trend is the same on arm an aarch64 for emulated runs, and I confirmed earlier that the results on real hardware are comparable to what we get in qemu.
