MacOS - Implement remaining portions for native ARM64 #15992
Conversation
|
Why is this in arch64 code? Isn't there already CPU detection for Mac? |
X86 uses CPU registers that are accessible in el0/ring3. Arm uses registers only accessible by the kernel at el1. On x86 builds apple just reports "virtual apple" as the cpu. |
| @@ -4,6 +4,10 @@ | |||
| #include <thread> | |||
| #include <map> | |||
|
|
|||
| #if defined(__APPLE__) | |||
There was a problem hiding this comment.
Could this path also be used for FREEBSD?
There was a problem hiding this comment.
Maybe, but I have no idea if they use a different path. I have 0 experience with BSD unfortunately, we'll have to rely on their community to help fill in the missing pieces.
kd-11
changed the title
| { | ||
| static std::mutex mcommit_lock; | ||
| std::lock_guard lock(mcommit_lock); | ||
| utils::memory_commit(pointer + olda, newa - olda, Prot); |
There was a problem hiding this comment.
hmm, this should somehow be implemented insidr memory_commit
edit: I'll add locks by pointer hash or something
There was a problem hiding this comment.
Beware of overlapping regions as well, though I doubt that's a problem in practice.
This operation is pretty quick so I haven't experienced any slowdown. It can also be copied as-is into the memory_commit function.
With these changes, LLVM is now usable on the M-series apple chips natively without going through rosetta. Same caveats as linux apply - not all games will work right now, but most of them should be fine, including AAA titles.