Commit Graph

241 Commits

Author SHA1 Message Date
Victor Stinner b298b395e8
gh-108765: Cleanup #include in Python/*.c files (#108977)
Mention one symbol imported by each #include.
2023-09-06 15:56:08 +02:00
Guido van Rossum 59e46932c8
gh-108488: Initialize JUMP_BACKWARD cache to 0, not 17 (#108591)
This mis-initialization caused the executor optimization to kick in sooner than intended. It also set the lower 4 bits of the counter to `1` -- those bits are supposed to be reserved (the actual counter is in the upper 12 bits).
2023-08-29 18:14:56 +00:00
Dong-hee Na 66b4d9c9f0
gh-107265: Revert "Ensure _PyCode_Quicken does not handle ENTER_EXECUTOR" (#108485)
This reverts commit d6ac5c7b10.

Reason: the assert we just added could be triggered (see issue).
2023-08-25 15:23:39 +00:00
Dong-hee Na d6ac5c7b10
gh-107265: Ensure _PyCode_Quicken does not handle ENTER_EXECUTOR (gh-108460) 2023-08-24 23:03:26 +00:00
Irit Katriel 72119d16a5
gh-105481: remove regen-opcode. Generated _PyOpcode_Caches in regen-cases. (#108367) 2023-08-23 18:39:00 +01:00
Brandt Bucher 326f0ba1c5
GH-106485: Dematerialize instance dictionaries when possible (GH-106539) 2023-08-09 19:14:50 +00:00
Brandt Bucher a9caf9cf90
GH-105848: Simplify the arrangement of CALL's stack (GH-107788) 2023-08-09 18:19:39 +00:00
Brandt Bucher ea72c6fe3b
GH-107596: Specialize str[int] (GH-107597) 2023-08-08 13:42:43 -07:00
Mark Shannon 2ba7c7f7b1
Add some GC stats to Py_STATS (GH-107581) 2023-08-04 10:34:23 +01:00
Victor Stinner 1a3faba9f1
gh-106869: Use new PyMemberDef constant names (#106871)
* Remove '#include "structmember.h"'.
* If needed, add <stddef.h> to get offsetof() function.
* Update Parser/asdl_c.py to regenerate Python/Python-ast.c.
* Replace:

  * T_SHORT => Py_T_SHORT
  * T_INT => Py_T_INT
  * T_LONG => Py_T_LONG
  * T_FLOAT => Py_T_FLOAT
  * T_DOUBLE => Py_T_DOUBLE
  * T_STRING => Py_T_STRING
  * T_OBJECT => _Py_T_OBJECT
  * T_CHAR => Py_T_CHAR
  * T_BYTE => Py_T_BYTE
  * T_UBYTE => Py_T_UBYTE
  * T_USHORT => Py_T_USHORT
  * T_UINT => Py_T_UINT
  * T_ULONG => Py_T_ULONG
  * T_STRING_INPLACE => Py_T_STRING_INPLACE
  * T_BOOL => Py_T_BOOL
  * T_OBJECT_EX => Py_T_OBJECT_EX
  * T_LONGLONG => Py_T_LONGLONG
  * T_ULONGLONG => Py_T_ULONGLONG
  * T_PYSSIZET => Py_T_PYSSIZET
  * T_NONE => _Py_T_NONE
  * READONLY => Py_READONLY
  * PY_AUDIT_READ => Py_AUDIT_READ
  * READ_RESTRICTED => Py_AUDIT_READ
  * PY_WRITE_RESTRICTED => _Py_WRITE_RESTRICTED
  * RESTRICTED => (READ_RESTRICTED | _Py_WRITE_RESTRICTED)
2023-07-25 15:28:30 +02:00
Victor Stinner ee15844db8
gh-106320: Move _PyMethodWrapper_Type to internal C API (#107064) 2023-07-22 20:57:59 +00:00
Mark Shannon 0c90e75610
GH-100288: Specialize LOAD_ATTR for simple class attributes. (#105990)
* Add two more specializations of LOAD_ATTR.
2023-07-10 11:40:35 +01:00
Guido van Rossum 17af98227f
gh-106320: Fix specialize.c compilation by including pycore_pylifecycle.h (#106434)
Compilation of Python/specialize.c was broken on macOS for me by gh-106400.
2023-07-04 21:42:12 +00:00
Brandt Bucher 7b2d94d875
GH-106008: Make implicit boolean conversions explicit (GH-106003) 2023-06-29 13:49:54 -07:00
Guido van Rossum 51fc725117
gh-104584: Baby steps towards generating and executing traces (#105924)
Added a new, experimental, tracing optimizer and interpreter (a.k.a. "tier 2"). This currently pessimizes, so don't use yet -- this is infrastructure so we can experiment with optimizing passes. To enable it, pass ``-Xuops`` or set ``PYTHONUOPS=1``. To get debug output, set ``PYTHONUOPSDEBUG=N`` where ``N`` is a debug level (0-4, where 0 is no debug output and 4 is excessively verbose).

All of this code is likely to change dramatically before the 3.13 feature freeze. But this is a first step.
2023-06-26 19:02:57 -07:00
Mark Shannon 04492cbc9a
GH-91095: Specialize calls to normal Python classes. (GH-99331) 2023-06-22 09:48:19 +01:00
Brandt Bucher 2beab5bdef
GH-105840: Fix assertion failures when specializing calls with too many __defaults__ (GH-105847) 2023-06-16 11:01:15 -07:00
Mark Shannon e830289c52
GH-105229: Remove remaining two-codeunit superinstructions (GH-105326)
* Remove LOAD_CONST__LOAD_FAST and LOAD_FAST__LOAD_CONST superinstructions.
2023-06-08 12:35:34 +01:00
Mark Shannon 0689340366
GH-105229: Replace some superinstructions with single instruction equivalent. (GH-105230) 2023-06-05 11:07:04 +01:00
Brandt Bucher 1eb950ca55
GH-104405: Add missing PEP 523 checks (GH-104406) 2023-05-12 22:23:13 +00:00
Carl Meyer 77262458fe
gh-87729: improve hit rate of LOAD_SUPER_ATTR specialization (#104270) 2023-05-11 08:08:13 -06:00
penguin_wwy 373bca0cc5
GH-102181: Improve specialization stats for SEND (GH-102182) 2023-05-10 22:40:59 +00:00
Carl Meyer afe7703744
gh-104184: fix building --with-pydebug --enable-pystats (#104217) 2023-05-09 08:53:19 -06:00
Mark Shannon 738c226786
GH-103082: Code cleanup in instrumentation code (#103474) 2023-04-29 04:51:55 +00:00
Carl Meyer ebf97c50f2
gh-103978: avoid using 'class' as an identifier (#103979) 2023-04-28 19:20:50 +00:00
Jelle Zijlstra 6c4124d11a
gh-103879: Fix refleak in super specialization (#103882) 2023-04-26 08:50:19 -07:00
Carl Meyer ef25febcf2
gh-87729: specialize LOAD_SUPER_ATTR_METHOD (#103809) 2023-04-25 17:45:51 +00:00
Mark Shannon 411b169281
GH-103082: Implementation of PEP 669: Low Impact Monitoring for CPython (GH-103083)
* The majority of the monitoring code is in instrumentation.c

* The new instrumentation bytecodes are in bytecodes.c

* legacy_tracing.c adapts the new API to the old sys.setrace and sys.setprofile APIs
2023-04-12 12:04:55 +01:00
Brandt Bucher b4978ff872
GH-88691: Shrink the CALL caches (GH-103230) 2023-04-05 14:15:49 -07:00
Brandt Bucher 121057aa36
GH-89987: Shrink the BINARY_SUBSCR caches (GH-103022) 2023-03-29 15:53:30 -07:00
Brandt Bucher 0444ae2487
GH-100982: Break up COMPARE_AND_BRANCH (GH-102801) 2023-03-23 15:25:09 -07:00
Mark Shannon 7559f5fda9
GH-101291: Rearrange the size bits in PyLongObject (GH-102464)
* Eliminate all remaining uses of Py_SIZE and Py_SET_SIZE on PyLongObject, adding asserts.

* Change layout of size/sign bits in longobject to support future addition of immortal ints and tagged medium ints.

* Add functions to hide some internals of long object, and for setting sign and digit count.

* Replace uses of IS_MEDIUM_VALUE macro with _PyLong_IsCompact().
2023-03-22 14:49:51 +00:00
Brandt Bucher 08b67fb34f
GH-90997: Shrink the LOAD_GLOBAL caches (#102569) 2023-03-10 17:01:16 -08:00
Eric Snow 5e5acd291f
gh-100227: Move next_keys_version to PyInterpreterState (gh-102335)
https://github.com/python/cpython/issues/100227
2023-03-08 18:04:16 -07:00
Mark Shannon 7c106a443f
GH-100982: Restrict `FOR_ITER_RANGE` to a single instruction to allow instrumentation. (GH-101985) 2023-02-22 11:11:57 +00:00
Steve Dower a99eb5cd99
gh-101907: Stop using `_Py_OPCODE` and `_Py_OPARG` macros (GH-101912)
* gh-101907: Removes use of non-standard C++ extension from Include/cpython/code.h

* Make cases_generator correct on Windows
2023-02-20 14:56:48 +00:00
Mark Shannon 160f2fe2b9
GH-87849: Simplify stack effect of SEND and specialize it for generators and coroutines. (GH-101788) 2023-02-13 11:24:55 +00:00
Brandt Bucher 76efcb4093
GH-100288: Skip extra work when failing to specialize LOAD_ATTR (GH-101354) 2023-01-31 13:28:32 -08:00
Mark Shannon c1b1f51cd1
GH-101291: Refactor the `PyLongObject` struct into object header and PyLongValue struct. (GH-101292) 2023-01-30 10:03:04 +00:00
Irit Katriel e9ccfe4a63
gh-100712: make it possible to disable specialization (for debugging) (#100713) 2023-01-19 18:14:55 +00:00
Mark Shannon 7b14c2ef19
GH-100982: Add `COMPARE_AND_BRANCH` instruction (GH-100983) 2023-01-16 12:35:21 +00:00
Mark Shannon 6e4e14d98f
GH-100923: Embed jump mask in `COMPARE_OP` oparg (GH-100924) 2023-01-11 20:40:43 +00:00
Mark Shannon f20c553a45
GH-100288: Remove LOAD_ATTR_METHOD_WITH_DICT instruction. (GH-100753) 2023-01-05 12:20:09 +00:00
Ken Jin 36d358348d
Revert "gh-100288: Specialise LOAD_ATTR_METHOD for managed dictionaries (GH-100289)" (#100468)
This reverts commit c3c7848a48.
2022-12-24 01:48:43 +08:00
Ken Jin c3c7848a48
gh-100288: Specialise LOAD_ATTR_METHOD for managed dictionaries (GH-100289) 2022-12-24 00:26:42 +08:00
Irit Katriel 2659036c75
GH-100459: fix copy-paste errors in specialization stats (GH-100460) 2022-12-23 14:42:24 +00:00
penguin_wwy a02161286a
GH-99770: Make the correct call specialization fail kind show up in the stats (GH-99771) 2022-12-22 12:44:59 +00:00
Dennis Sweeney c18d831188
gh-100188: Reduce misses in BINARY_SUBSCR_(LIST/TUPLE)_INT (#100189)
Don't specialize if the index is negative.
2022-12-20 15:46:16 -05:00
Brandt Bucher 9076455d1b
GH-90043: Handle NaNs in COMPARE_OP_FLOAT_JUMP (GH-100278) 2022-12-16 10:18:31 -08:00
Mark Shannon d4052d835b
Improve stats presentation for calls. (GH-100274) 2022-12-16 15:43:04 +00:00
Mark Shannon 289c1126dd
Better stats for `LOAD_ATTR` and `STORE_ATTR` (GH-100295)
* Don't attempt to specialize for LOAD_ATTR on instance if class has attribute

* Improvement to LOAD_ATTR and STORE_ATTR specialization stats.
2022-12-16 15:41:23 +00:00
Mark Shannon 48e352a241
Move stats for the method cache into the `Py_STAT` machinery (GH-100255) 2022-12-15 09:45:03 +00:00
Carl Meyer bdd86741be
GH-100222: fix typo _py_set_opocde -> _py_set_opcode (GH-100259)
Typo introduced in #100223.

Automerge-Triggered-By: GH:brandtbucher
2022-12-14 16:39:00 -08:00
Mark Shannon 5693f45b19
Assorted minor fixes for specialization stats. (GH-100219) 2022-12-14 15:50:02 +00:00
Mark Shannon 6997e77bdf
GH-100222: Redefine _Py_CODEUNIT as a union to clarify structure of code unit. (GH-100223) 2022-12-14 11:12:53 +00:00
Michael Droettboom 1583c6e326
GH-100143: Improve collecting pystats for parts of runs (GH-100144)
* pystats off by default

* Add -Xpystats flag

* Always dump pystats, even if turned off
2022-12-12 14:50:43 +00:00
Ken Jin 748c6c0921
GH-100110: Specialize FOR_ITER for tuples (GH-100109)
* Specialize FOR_ITER for tuples
2022-12-09 10:27:01 +00:00
Brandt Bucher b629fdd88a
GH-99298: Clean up attribute specializations (GH-99398) 2022-11-17 15:09:18 -08:00
Brandt Bucher 8555dee5ae
GH-98686: Get rid of BINARY_OP_GENERIC and COMPARE_OP_GENERIC (GH-99399) 2022-11-17 11:36:57 -08:00
Brandt Bucher 9d69284169
GH-99257: Check the owner's type when specializing slots (GH-99258) 2022-11-10 11:50:34 +00:00
Brandt Bucher c7f5708714
GH-98686: Get rid of "adaptive" and "quick" instructions (GH-99182) 2022-11-09 10:50:09 -08:00
Mark Shannon 4a1c58d504
GH-96793: Specialize FOR_ITER for generators. (GH-98772) 2022-11-07 14:49:51 +00:00
Brandt Bucher 276d77724f
GH-98686: Quicken everything (GH-98687) 2022-11-02 10:42:57 -07:00
Mark Shannon 22863df7ca
GH-96793: Change `FOR_ITER` to not pop the iterator on exhaustion. (GH-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).
2022-10-27 11:55:03 +01:00
adphrost a41ed975e8
GH-91049: Introduce set vectorcall field API for PyFunctionObject (GH-92257)
Co-authored-by: Andrew Frost <adfrost@fb.com>
Co-authored-by: Itamar Ostricher <itamarost@gmail.com>
2022-09-15 16:42:37 +01:00
Brandt Bucher b4954b1a9e
GH-90230: Fix warnings and failures with --enable-pystats (GH-96622) 2022-09-09 14:42:29 -07:00
Brandt Bucher cd0ff9bd14
GH-93911: Fix `LOAD_ATTR_PROPERTY` caches (GH-96519) 2022-09-06 12:11:38 +01:00
Irit Katriel 4c72517cad
gh-93554: Conditional jump opcodes only jump forward (GH-96318) 2022-09-01 21:36:47 +01:00
Matthias Görgens 4a6fa89465
Remove dead code in _PyDict_GetItemHint and rename to _PyDict_LookupIndex (GH-95948) 2022-08-18 10:19:21 +01:00
Ken Jin 7276ca25f5
GH-93911: Specialize `LOAD_ATTR` for custom `__getattribute__` (GH-93988) 2022-08-17 12:37:07 +01:00
Mark Shannon de388c0a7b
GH-95245: Store object values and dict pointers in single tagged pointer. (GH-95278) 2022-08-01 14:34:54 +01:00
Brandt Bucher e402b26b7f
GH-95113: Don't use EXTENDED_ARG_QUICK in unquickened code (GH-95121) 2022-07-22 11:04:20 -07:00
Brandt Bucher daf68ba92f
GH-94822: Don't specialize when metaclasses are involved (GH-94892) 2022-07-18 10:10:22 -07:00
Mark Shannon c0453a40fa
GH-94163: Add BINARY_SLICE and STORE_SLICE instructions. (GH-94168) 2022-06-27 12:24:23 +01:00
Mark Shannon 6f8875eba3
GH-93841: Allow stats to be turned on and off, cleared and dumped at runtime. (GH-93843) 2022-06-21 15:40:54 +01:00
Dennis Sweeney 5fcfdd87c9
GH-91432: Specialize FOR_ITER (GH-91713)
* Adds FOR_ITER_LIST and FOR_ITER_RANGE specializations.

* Adds _PyLong_AssignValue() internal function to avoid temporary boxing of ints.
2022-06-21 11:19:26 +01:00
Christian Heimes 77c839c98f
gh-94021: Address unreachable code warning in specialize code (GH-94022) 2022-06-20 23:34:11 -07:00
Ken Jin a51742ab82
gh-93911: Specialize `LOAD_ATTR_PROPERTY` (GH-93912) 2022-06-17 23:13:17 +08:00
Ken Jin ab45c1dde0
Fix BINARY_SUBSCR_GETITEM stats (GH-93903) 2022-06-16 15:02:07 +01:00
Mark Shannon 6b330002b1
Rename 'LOAD_METHOD' specialization stat consts to 'ATTR'. (GH-93812) 2022-06-14 16:44:01 +01:00
Mark Shannon ef6e44d392
Remove LOAD_METHOD stats. (GH-93807) 2022-06-14 15:03:15 +01:00
Ken Jin b083450f88
GH-93429: Merge `LOAD_METHOD` back into `LOAD_ATTR` (GH-93430) 2022-06-14 11:36:22 +01:00
Dennis Sweeney c5d0517ea4
Add more FOR_ITER specialization stats (GH-32151) 2022-06-13 01:40:54 -04:00
Mark Shannon f012df706c
Shrink the LOAD_METHOD cache by one codeunit. (#93537) 2022-06-07 10:28:53 +01:00
Mark Shannon eb618d5ff0
GH-93354: Use exponential backoff to avoid excessive specialization attempts. (GH-93355) 2022-05-31 11:58:26 +01:00
Mark Shannon bbcf42449e
GH-90230: Add stats to breakdown the origin of calls to `PyEval_EvalFrame` (GH-93284) 2022-05-27 16:31:41 +01:00
Ken Jin 5e6e5b98a8
gh-92777: Add LOAD_METHOD_LAZY_DICT (GH-92778) 2022-05-25 14:06:15 +01:00
Mark Shannon e48ac9c100
GH-90690: Remove `PRECALL` instruction (GH-92925) 2022-05-19 11:05:26 +01:00
Mark Shannon a4460f2eb8
Split refcount stats into 'interpreter' and 'non-interpreter' (GH-92919) 2022-05-18 14:38:43 +01:00
Mark Shannon fa2b8b75eb
Improve object stats (#92845)
* Add incref/decref stats

* Show ratios for allocation in summary
2022-05-16 14:35:11 +01:00
Mark Shannon f8a2fab212
GH-92239: Make sure that PEP 523 is supported, even when specializing first. (GH-92245) 2022-05-04 09:31:21 -06:00
Mark Shannon 836b17c9c3
Add more stats for freelist use and allocations. (GH-92211) 2022-05-03 16:40:24 -06:00
Victor Stinner 64a54e511d
gh-91719: Add pycore_opcode.h internal header file (#91906)
Move the following API from Include/opcode.h (public C API) to a new
Include/internal/pycore_opcode.h header file (internal C API):

* EXTRA_CASES
* _PyOpcode_Caches
* _PyOpcode_Deopt
* _PyOpcode_Jump
* _PyOpcode_OpName
* _PyOpcode_RelativeJump
2022-04-26 00:14:30 +02:00
Irit Katriel dd207a6ac5
bpo-47120: make POP_JUMP_IF_TRUE/FALSE/NONE/NOT_NONE relative (GH-32400) 2022-04-11 10:40:24 +01:00
Dennis Sweeney 6c6e0408a6
bpo-47009: Let PRECALL_NO_KW_LIST_APPEND do its own POP_TOP (GH-32239) 2022-04-05 11:18:30 +01:00
Irit Katriel a00518d9ad
bpo-47120: Replace the JUMP_ABSOLUTE opcode by the relative JUMP_BACKWARD (GH-32115) 2022-03-31 14:14:15 +01:00
Mark Shannon 04acfa94bb
Merge deoptimization blocks in interpreter (GH-32155) 2022-03-30 13:11:33 +01:00
Kumar Aditya 58448cbd96
bpo-47127: Specialize calls for fastcall c methods with keywords (GH-32125)
* add PRECALL_METHOD_DESCRIPTOR_FAST_WITH_KEYWORDS
2022-03-28 03:53:25 +08:00
Dennis Sweeney cca43b7d64
bpo-47053: Reduce deoptimization in BINARY_OP_INPLACE_ADD_UNICODE (GH-31318)
* Don't deopt if refcounts are too big

* Detect more at specialization time
2022-03-25 16:13:19 +00:00
Brandt Bucher 2bde6827ea
bpo-46841: Quicken code in-place (GH-31888)
* Moves the bytecode to the end of the corresponding PyCodeObject, and quickens it in-place.

* Removes the almost-always-unused co_varnames, co_freevars, and co_cellvars member caches

* _PyOpcode_Deopt is a new mapping from all opcodes to their un-quickened forms.

* _PyOpcode_InlineCacheEntries is renamed to _PyOpcode_Caches

* _Py_IncrementCountAndMaybeQuicken is renamed to _PyCode_Warmup

* _Py_Quicken is renamed to _PyCode_Quicken

* _co_quickened is renamed to _co_code_adaptive (and is now a read-only memoryview).

* Do not emit unused nonzero opargs anymore in the compiler.
2022-03-21 11:11:17 +00:00