Enable specialization of iteration over generators, `for gen():`, and awaiting coroutines, `await coro()` #96793

markshannon · 2022-09-13T11:51:35Z

We want to be able to specialize iteration over generators and awaiting coroutines. Both for PEP 659 and for higher tier optimizers.

For the theory see faster-cpython/ideas#457

To do this we need:

Shim frames to keep the various returns simple GH-96421: Insert shim frame on entry to interpreter #96319
Modify FOR_ITER to leave a value on the stack when exiting.
Add a GEN_RETURN_VALUE instruction to allow the return instruction to clean up the frame.
Implement specialization of FOR_ITER for generators
Implement specialization of SEND for coroutines

The text was updated successfully, but these errors were encountered:

…H-96801) Change FOR_ITER to have the same stack effect regardless of whether it branches or not. Performance is unchanged as FOR_ITER (and specialized forms jump over the cleanup code).

markshannon added the performance Performance or resource usage label Sep 13, 2022

markshannon mentioned this issue Sep 13, 2022

GH-96793: Change FOR_ITER to not pop the iterator on exhaustion. #96801

Merged

Enable specialization of iteration over generators, for gen():, and awaiting coroutines, await coro() #96793

Enable specialization of iteration over generators, for gen():, and awaiting coroutines, await coro() #96793

Comments

markshannon commented Sep 13, 2022 • edited

Enable specialization of iteration over generators, `for gen():`, and awaiting coroutines, `await coro()` #96793

Enable specialization of iteration over generators, `for gen():`, and awaiting coroutines, `await coro()` #96793

markshannon commented Sep 13, 2022 •

edited