Apple Silicon has special support for running ARM code with the "Total Store Order" memory model of x86.
JIT-compiling emulators for other ARM processors otherwise need to be stuff the code full of memory fence instructions to be able to utilise multiple cores properly. Optimising those away can be hard.
JIT-compiling emulators for other ARM processors otherwise need to be stuff the code full of memory fence instructions to be able to utilise multiple cores properly. Optimising those away can be hard.