Implement (some/all of) the things Rosetta 2 does to achieve high x86_64 performance on Apple Silicon

Hi all,
As is well known Rosetta 2 runs x86_64 code on Apple Silicon typically 4-5X faster than QEMU/UTM. These are big gains. This speed-up is due primarily to 3 things that QEMU/UTM does not do on macOS:

1. Use of TSO mode (removes the need for memory fences)
2. Return address prediction (rewriting CALL/RET with BL/RET with some extra housekeeping)
3. Use of Apple Silicon ARM extensions to handle x86_64 flag setting directly (SETF8, SETF16, AXFLAG, XAFLAG)

#1 probably gives the largest gains and can be enabled from userland with the com.apple.private.oahd entitlement. Yes, this couldn't be done with the App Store version of UTM, but shouldn't cause any difficulties for the direct-download version.

#2 and #3 can be achieved without any special entitlements.

QEMU may or may not have interest in implementing these since they are Apple Silicon specific. Has UTM considered these for its QEMU fork? One or more of these changes could lead to large gains.