GH-132532: Add new DSL macros to better declare semantics of exits at ends of instructions/uops. #137098
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ics of exits at ends of instructions/uops.
brandtbucher
approved these changes
Aug 7, 2025
Comment on lines
+165
to
+170
| _Py_CHECK_EMSCRIPTEN_SIGNALS_PERIODICALLY(); | ||
| QSBR_QUIESCENT_STATE(tstate); | ||
| if (_Py_atomic_load_uintptr_relaxed(&tstate->eval_breaker) & _PY_EVAL_EVENTS_MASK) { | ||
| int err = _Py_HandlePending(tstate); | ||
| ERROR_IF(err != 0); | ||
| } |
There was a problem hiding this comment.
This logic is now repeated 3 times, and two of them are in identical opcode implementations. Maybe it should be broken out into a static inline helper function?
| @@ -5189,32 +5198,28 @@ dummy_func( | |||
| op (_GUARD_IS_TRUE_POP, (flag -- )) { | |||
| int is_true = PyStackRef_IsTrue(flag); | |||
| DEAD(flag); | |||
| SYNC_SP(); | |||
| EXIT_IF(!is_true); | |||
| EXIT_IF_AFTER(!is_true); | |||
There was a problem hiding this comment.
The wording makes this a bit hard to parse: "exit if after not is true"? Maybe use "at end" for consistency with _CHECK_PERIODIC_AT_END, if they mean the same thing? AT_END_EXIT_IF, maybe?
| @@ -5380,6 +5385,11 @@ dummy_func( | |||
| GOTO_TIER_ONE(_PyFrame_GetBytecode(frame) + CURRENT_TARGET()); | |||
| } | |||
|
|
|||
| tier2 op(_PERIODIC, (--)) { | |||
There was a problem hiding this comment.
I get that this is "check periodic without the check", but maybe we could find a better name here? _HANDLE_PENDING_AND_DEOPT?
| @@ -5391,11 +5401,11 @@ dummy_func( | |||
| * ENTER_EXECUTOR will not re-enter tier 2 with the eval breaker set. */ | |||
| tier2 op(_TIER2_RESUME_CHECK, (--)) { | |||
| #if defined(__EMSCRIPTEN__) | |||
| DEOPT_IF(_Py_emscripten_signal_clock == 0); | |||
| PERIODIC_IF(_Py_emscripten_signal_clock == 0); | |||
There was a problem hiding this comment.
Same here... Maybe HANDLE_PENDING_AND_DEOPT_IF?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR changes adds the ability to de-opt at the end an instruction.
Exits from tier 2 jump to the next instruction in tier 1.
In the case of deopting due to the eval breaker being set,
_Py_HandlePendingis called before resuming tier 1.I am taking this approach instead of checking the eval breaker before each call, as this does not cause any observable behavior change. For comparison see https://github.com/python/cpython/compare/main...faster-cpython:cpython:check-periodic-before-call?expand=1, noting the necessary changes to tests and threading module.
Although this approach adds a bit more complexity to the optimizer, it doesn't need any changes to the library or tests.