Skip to content

Support RETURN from tracing through a function #107925

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
wants to merge 35 commits into from
Closed

Support RETURN from tracing through a function #107925

wants to merge 35 commits into from

Conversation

gvanrossum
Copy link
Member

@gvanrossum gvanrossum commented Aug 14, 2023 •
edited
Loading

This depends on gh-107793.

gvanrossum added 17 commits August 9, 2023 11:41
@gvanrossum
Split CALL_PY_EXACT_ARGS into uops
56133bb 
This is only the first step for doing `CALL` in Tier 2.
The next step involves tracing into the called code object.
After that we'll have to do the remaining `CALL` specialization.
Finally we'll have to tweak various things like `KW_NAMES`,
and possibly move the `NULL` (for method calls) *above* the callable.
But those are things for future PRs.

Note: this moves setting `frame->return_offset` directly in front of
`DISPATCH_INLINED()`, to make it easier to move it into `_PUSH_FRAME`.
@gvanrossum
Fix merge so it works again (I think)
907ff95
@gvanrossum
Split into finer-grained uops
2c6be6d
@gvanrossum
Fix type error in stacking.py
6d78ff2
@gvanrossum
Add test
0d8e66c
@gvanrossum
Add comment explaining _PUSH_FRAME's unused output effect
b75f30e
@gvanrossum
Make PUSH_FRAME special case a little less myterious
61c2822
@gvanrossum
Rename Instruction.write to write_case_body
f73ea90
@gvanrossum
Move next_instr update to a more logical place
12910fc
@gvanrossum
Don't recompute macro cache offset
2fafa2c
@gvanrossum
Fold and refactor long line in stacking.py
2717b07
@gvanrossum
Fold long lines in generate_cases.py
e487908
@gvanrossum
Don't emit static assert to executor cases
1d549af
@gvanrossum
Factor away write_case_body (formerly Instruction.write)
f40fb1f
@gvanrossum
Fold long lines
4f6f8f8
@gvanrossum
Make less of a special case of _PUSH_FRAME
6facc8d 
Instead, the special case is an opcode using SAVE_FRAME_STATE().
Introducing #if TIER_ONE and #if TIER_TWO so we can implement
_PUSH_FRAME differently for both tiers.
@gvanrossum
Stop special-casing _PUSH_FRAME altogether
94630d4 
Instead, we special-case SAVE_IP:
- Its Tier 2 expansion sets oparg to the instruction offset
- In Tier 1 it is a no-op (and skipped if present in a macro)
This was referenced Aug 14, 2023
Closed
Closed
@gvanrossum
Copy link
Member Author

gvanrossum commented Aug 14, 2023 •
edited
Loading

Running tests with -Xuops seems to indicate that my translation of RETURN_CONST into LOAD_CONST + _POP_FRAME doesn't work -- it requires an extra stack slot which isn't always available. Back to the drawing board...

@gvanrossum
Copy link
Member Author

Running tests with -Xuops seems to indicate that my translation of RETURN_CONST into LOAD_CONST + _POP_FRAME doesn't work -- it requires an extra stack slot which isn't always available. Back to the drawing board...

This should be easy -- require that every function has at least one stack slot. The compiler guarantees that there is nothing on the stack below the return value, so RETURN_CONST would always start with an empty stack.

@gvanrossum gvanrossum mentioned this pull request Aug 14, 2023
Closed
@gvanrossum gvanrossum mentioned this pull request Aug 15, 2023
Closed
@gvanrossum
Add function-by-version cache
337a2eb
@gvanrossum
Trace into function calls
0218222 
- This uses the function-by-version cache I just added
- There's not yet a way to trace back via `RETURN_VALUE`
@gvanrossum
Make RESUME a viable uop
ab74ef0
@gvanrossum
Clear func_version_cache in interpreter_clear()
7e6ef78 
This should fix leaks and hopefully most failing tests.
@gvanrossum
Add a small essay on function and code versions
1200014
@gvanrossum
Move function cache clearing earlier in finalization
ece96f4
@gvanrossum
Cache borrowed references, cleared in func_dealloc
b408b2b
@gvanrossum
Correctly set instr_fmt metadata for macros
00ae0fa 
Macros not using oparg should have a format starting with IX, not IB.
This corrects the metadata for a bunch of specializations, e.g.
BINARY_OP_MULTIPLY_INT (which never looks at oparg).
@gvanrossum
Split RETURN_{VALUE,CONST} into uops (mostly works)
fe843d8
@gvanrossum
Change LLTRACE debug to trigger on PYTHONUOPSDEBUG >= 5, not 4
5bf745c
@gvanrossum
Handle _POP_FRAME in superblock properly
1e55726
@gvanrossum
Handle trace stack underflow
c632262
@gvanrossum
Add _Py_LeaveRecursiveCallPy to _POP_FRAME
b59a3af
@gvanrossum
Ensure co_stacksize >= 1, else RETURN_CONST may crash
158e27b
@gvanrossum
Fix failing test_code (co_stacksize is never 0)
8d95157
@gvanrossum
Don't trace into functions if func_version != co_version
d90ecaf 
Seems to fix test_opcache, playing games with func.__code__ = func.__code__.replace().
@gvanrossum gvanrossum mentioned this pull request Aug 16, 2023
Merged
4 tasks
@gvanrossum gvanrossum marked this pull request as ready for review August 16, 2023 04:30
@gvanrossum gvanrossum mentioned this pull request Aug 16, 2023
Merged
@gvanrossum gvanrossum closed this Aug 16, 2023
@gvanrossum gvanrossum deleted the return-uops branch August 17, 2023 00:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@markshannon markshannon Awaiting requested review from markshannon markshannon is a code owner

Assignees
No one assigned
Labels
awaiting core review skip news
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@gvanrossum @bedevere-bot