Open diegorusso opened 2 weeks ago
The proposal is an internal implementation of the JIT and it transparent to the user.
I have already discussed this feature proposal on Discourse
Not on Discourse, I've had extensive discussion about this with @brandtbucher and created a prototype that we benchmarked using the faster-cpython infrastructure. The full results are here: https://github.com/faster-cpython/benchmarking-public/tree/main/results/bm-20241018-3.14.0a1+-f74cd79-JIT
On macOS on Apple Silicon there is an overall 1.4% speed improvement by replacing the mprotect with pthread_jit_write_protect_np.
mprotect
pthread_jit_write_protect_np
I'll leave this open in case we want to explore the pthread_jit_write_with_callback_np API: https://developer.apple.com/documentation/apple-silicon/porting-just-in-time-compilers-to-apple-silicon
pthread_jit_write_with_callback_np
Feature or enhancement
Proposal:
The proposal is an internal implementation of the JIT and it transparent to the user.
Has this already been discussed elsewhere?
I have already discussed this feature proposal on Discourse
Links to previous discussion of this feature:
Not on Discourse, I've had extensive discussion about this with @brandtbucher and created a prototype that we benchmarked using the faster-cpython infrastructure. The full results are here: https://github.com/faster-cpython/benchmarking-public/tree/main/results/bm-20241018-3.14.0a1+-f74cd79-JIT
On macOS on Apple Silicon there is an overall 1.4% speed improvement by replacing the
mprotect
withpthread_jit_write_protect_np
.Linked PRs