Mark Shannon
70bd1c2dd2
GH-143732: SEND specialization (GH-148963)
...
* SEND specialization. Adds 2 new specialized instructions:
* SEND_VIRTUAL: for sends to virtual iterators e.g lists and tuples
* SEND_ASYNC_GEN: for sends to async generators
Tweak FOR_ITER_VIRTUAL so that SEND_VIRTUAL and FOR_ITER_VIRTUAL use equivalent guards
2026-05-05 15:19:16 +01:00
Pablo Galindo Salgado
9a268e3e33
gh-98894: Restore function entry/exit DTrace probes ( #142397 )
...
The function__entry and function__return probes stopped working in Python 3.11
when the interpreter was restructured around the new bytecode system. This change
restores these probes by adding DTRACE_FUNCTION_ENTRY() at the start_frame label
in bytecodes.c and DTRACE_FUNCTION_RETURN() in the RETURN_VALUE and YIELD_VALUE
instructions. The helper functions are defined in ceval.c and extract the
filename, function name, and line number from the frame before firing the probe.
This builds on the approach from https://github.com/python/cpython/pull/125019
but avoids modifying the JIT template since the JIT does not currently support
DTrace. The macros are conditionally compiled with WITH_DTRACE and are no-ops
otherwise. The tests have been updated to use modern opcode names (CALL, CALL_KW,
CALL_FUNCTION_EX) and a new bpftrace backend was added for Linux CI alongside
the existing SystemTap tests. Line probe tests were removed since that probe
was never restored after 3.11.
2026-05-05 00:29:55 +00:00
Neko Asakura
9846407eaf
gh-143732: add specialization for FOR_ITER (GH-148745)
2026-05-04 17:29:10 +01:00
Victor Stinner
952784af47
gh-137030: Fix YIELD_VALUE bytecode assertion ( #149184 )
...
Co-authored-by: Mark Shannon <mark@hotpy.org >
2026-05-04 18:09:57 +02:00
Kumar Aditya
5847931d11
gh-143732: allow dict subclasses to be specialized (GH-148128)
2026-05-04 09:39:03 +01:00
sobolevn
726a17e265
gh-149243: Check for recursion limits in CALL_ALLOC_AND_ENTER_INIT ( #149310 )
2026-05-03 13:26:05 +03:00
Neko Asakura
2ca6333065
gh-148380: remove all uses of _PyType_LookupByVersion in optimizer_bytecodes.c (GH-148394)
2026-05-02 15:45:38 +01:00
Neko Asakura
0102c1d9b9
gh-149204: add _RROT_3 uop to reduce stack moves (GH-149205)
2026-05-01 12:35:31 +01:00
Neko Asakura
2b6a13710f
gh-148211: decompose _SHUFFLE_3_LOAD_CONST_INLINE_BORROW in JIT (GH-148816)
2026-04-28 13:48:23 +01:00
Hai Zhu
618b726d68
gh-146073: Add fitness/exit quality mechanism for JIT trace frontend (GH-148089)
...
* Replaces ad-hoc logic for ending traces with a simple inequality: `fitness < exit_quality`
* Fitness starts high and is reduced for branches, backward edges, calls and trace length
* Exit quality reflect how good a spot that instruction is to end a trace. Closing a loop is very, specializable instructions are very low and the others in between.
2026-04-24 10:37:01 +01:00
Mark Shannon
600f4dbd54
GH-145668: Add FOR_ITER specialization for virtual iterators. Specialize GET_ITER. (GH-147967)
...
* Add FOR_ITER_VIRTUAL to specialize FOR_ITER for virtual iterators
* Add GET_ITER_SELF to specialize GET_ITER for iterators (including generators)
* Add GET_ITER_VIRTUAL to specialize GET_ITER for iterables as virtual iterators
* Add new (internal) _tp_iteritem function slot to PyTypeObject
* Put limited RESUME at start of genexpr for free-threading. Fix up exception handling in genexpr
2026-04-16 15:22:22 +01:00
Pieter Eendebak
1f6a09fb36
gh-100239: Specialize more binary operations using BINARY_OP_EXTEND (GH-128956)
2026-04-16 09:22:41 +01:00
Ken Jin
ab45919812
gh-148609: Remove unicode character in bytecodes.c (GH-148611)
...
Remove unicode character in bytecodes.c
2026-04-15 12:58:35 +00:00
Pieter Eendebak
5c3decad66
gh-146393: Use recorded type instead of instance in BINARY_OP ( #148569 )
2026-04-14 20:11:42 +00:00
Pieter Eendebak
95cbd4a232
gh-146393: Optimize float division operations by mutating uniquely-referenced operands in place (JIT only) (GH-146397)
...
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com >
2026-04-15 02:08:04 +08:00
Kumar Aditya
3cb7eaec85
gh-131798: constant fold special method lookups in JIT ( #148432 )
2026-04-14 16:02:23 +00:00
Kumar Aditya
1aa7e7ee6d
gh-gh-131798: optimize LOAD_ATTR_GETATTRIBUTE_OVERRIDDEN in the JIT ( #148555 )
2026-04-14 21:00:32 +05:30
Neko Asakura
52a7f1b7f8
gh-148510: restore func_version check in _LOAD_ATTR_PROPERTY_FRAME (GH-148528)
2026-04-14 22:44:39 +08:00
Hai Zhu
5ce0fe8b6c
gh-148378: Allow multiple consecutive recording ops per macro op (GH-148496)
2026-04-14 19:26:53 +08:00
Kumar Aditya
88e378cc1c
gh-131798: optimize through keyword and bound method calls in the JIT (GH-148466)
...
Co-authored-by: Ken Jin <kenjin4096@gmail.com >
2026-04-13 21:14:48 +08:00
Neko Asakura
63492628be
gh-148438: implement _RECORD_BOUND_METHOD in JIT (GH-148457)
2026-04-13 02:57:55 +08:00
Sacul
18d7d90ef9
gh-131798: Split _CHECK_AND_ALLOCATE_OBJECT into smaller uops (GH-148433)
...
Co-authored-by: Hai Zhu <haiizhu@outlook.com >
Co-authored-by: Ken Jin <kenjin4096@gmail.com >
2026-04-13 02:31:24 +08:00
Pieter Eendebak
1c89817f51
gh-148276: Optimize object creation and method calls in the JIT by resolving __init__ at trace optimization time (GH-148277)
...
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com >
Co-authored-by: Ken Jin <kenjin4096@gmail.com >
2026-04-11 22:22:42 +08:00
Donghee Na
a71b043356
gh-148171: Convert CALL_BUILTIN_CLASS to leave arguments on the stack (gh-148381)
2026-04-11 23:01:25 +09:00
Neko Asakura
9831dea3bf
gh-148211: decompose _POP_TWO/_POP_CALL(_ONE/_TWO) in JIT (GH-148377)
2026-04-11 20:46:56 +08:00
Neko Asakura
72006a71b2
gh-148211: decompose [_POP_TWO/_INSERT_2]_LOAD_CONST_INLINE_BORROW in JIT (GH-148357)
2026-04-11 18:27:51 +08:00
Sacul
e872c19922
gh-148171: Convert variadic argument opcodes to leave their arguments on the stack (CALL_BUILTIN_FAST_WITH_KEYWORDS) ( #148366 )
2026-04-11 13:27:24 +08:00
Kumar Aditya
2b439da972
gh-148171: convert more variadic uops to leave input on stack in JIT ( #148361 )
2026-04-11 10:29:38 +05:30
Kumar Aditya
8f17140fc1
gh-131798: split _CALL_BUILTIN_CLASS to smaller uops ( #148094 )
2026-04-10 17:28:20 +00:00
Ken Jin
266247c9a6
gh-148171: Convert CALL_BUILTIN_FAST to leave inputs on the stack for refcount elimination in JIT (GH-148172)
2026-04-10 23:11:18 +08:00
Neko Asakura
aea0b91d65
gh-148211: decompose [_POP_CALL_X/_SHUFFLE_2]_LOAD_CONST_INLINE_BORROW in JIT (GH-148313)
2026-04-10 21:57:01 +08:00
Neko Asakura
0f49232664
gh-148211: decompose _INSERT_1_LOAD_CONST_INLINE(_BORROW) in JIT (GH-148283)
2026-04-10 00:45:39 +08:00
Sacul
38d3aef375
gh-134584 : Optimize and eliminate redundant ref-counting for MAKE_FUNCTION in the JIT (GH-144963)
2026-04-09 22:22:53 +08:00
Kumar Aditya
458aca9237
gh-131798: fold super method lookups in JIT ( #148231 )
2026-04-09 13:25:01 +05:30
Sacul
bb03c8bd02
gh-145866: Convert _CALL_METHOD_DESCRIPTOR_NOARGS to leave its inputs on the stack to be cleaned up by _POP_TOP (GH-148227)
2026-04-08 23:21:37 +08:00
Neko Asakura
756358524e
gh-148235: remove duplicate uops _LOAD_CONST_UNDER_INLINE(_BORROW) in JIT (GH-148236)
2026-04-08 16:22:59 +08:00
Kumar Aditya
e7bf8eac0f
gh-131798: split recursion check to _CHECK_RECURSION_LIMIT and combine checks (GH-148070)
2026-04-04 17:23:03 +08:00
Kumar Aditya
7e275d4965
gh-131798: JIT inline function addresses of builtin methods ( #146906 )
2026-04-04 09:12:13 +05:30
Pieter Eendebak
48317feec8
gh-146640: Optimize int operations by mutating uniquely-referenced operands in place (JIT only) (GH-146641)
2026-04-03 23:23:04 +08:00
Kumar Aditya
8e9d21c64b
gh-146558: JIT optimize dict access for objects with known hash ( #146559 )
2026-03-30 14:23:29 +00:00
Serhiy Storchaka
6932c3ee6a
gh-145876: Do not mask KeyErrors raised during dictionary unpacking in call (GH-146472)
...
KeyErrors raised in keys() or __getitem__() during dictionary unpacking
in call (func(**mymapping)) are no longer masked by TypeError.
2026-03-29 11:58:52 +03:00
Neko Asakura
a492d9ff74
gh-145866: Convert LIST_EXTEND to leave its inputs on the stack to be cleaned up by _POP_TOP be cleaned up by _POP_TOP (GH-146383)
2026-03-28 01:04:17 +08:00
Kumar Aditya
bfdaa3c565
gh-131798: JIT: split call method and call builtin opcodes into smaller uops ( #146463 )
2026-03-26 14:20:57 +00:00
Serhiy Storchaka
0e543055b0
gh-145876: Do not mask AttributeErrors raised during dictionary unpacking (GH-145906)
...
AttributeErrors raised in keys() or __getitem__() during
dictionary unpacking ({**mymapping} or func(**mymapping)) are
no longer masked by TypeError.
2026-03-26 13:48:57 +00:00
Sacul
1516c26399
gh-145866: Convert DICT_UPDATE to leave its inputs on the stack to be cleaned up by _POP_TOP (GH-146190)
2026-03-26 19:38:44 +08:00
Kevin Huai
f5364ae750
gh-145866: Convert _CALL_INTRINSIC_2 to leave its inputs on the stack to be cleaned up by _POP_TOP (GH-146262)
2026-03-26 17:41:07 +08:00
Pieter Eendebak
951675c18a
gh-146306: JIT: Optimize float operations by mutating uniquely-referenced operands in place (GH-146307)
...
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com >
2026-03-24 21:16:02 +08:00
Neko Asakura
6d73bc2267
gh-145866: Convert DICT_MERGE to leave its inputs on the stack to be cleaned up by _POP_TOP (GH-146329)
2026-03-24 18:31:41 +08:00
reiden
e36f8db7e5
gh-143414: Implement unique reference tracking for JIT, optimize unpacking of such tuples (GH-144300)
...
Co-authored-by: Ken Jin <kenjin4096@gmail.com >
2026-03-23 00:57:23 +08:00
Mark Shannon
879c85f6e4
GH-145667: Merge GET_ITER and GET_YIELD_FROM_ITER (GH-146120)
...
* Merge GET_ITER and GET_YIELD_FROM_ITER. Modify SEND to make it a bit more like FOR_ITER
2026-03-21 10:48:13 +00:00