Rename private C API constants:
* Rename PY_MONITORING_UNGROUPED_EVENTS to _PY_MONITORING_UNGROUPED_EVENTS
* Rename PY_MONITORING_EVENTS to _PY_MONITORING_EVENTS
By turning `assert(kwnames == NULL)` into a macro that is not in the "forbidden" list, many instructions that formerly were skipped because they contained such an assert (but no other mention of `kwnames`) are now supported in Tier 2. This covers 10 instructions in total (all specializations of `CALL` that invoke some C code):
- `CALL_NO_KW_TYPE_1`
- `CALL_NO_KW_STR_1`
- `CALL_NO_KW_TUPLE_1`
- `CALL_NO_KW_BUILTIN_O`
- `CALL_NO_KW_BUILTIN_FAST`
- `CALL_NO_KW_LEN`
- `CALL_NO_KW_ISINSTANCE`
- `CALL_NO_KW_METHOD_DESCRIPTOR_O`
- `CALL_NO_KW_METHOD_DESCRIPTOR_NOARGS`
- `CALL_NO_KW_METHOD_DESCRIPTOR_FAST`
This moves EXIT_TRACE, SAVE_IP, JUMP_TO_TOP, and
_POP_JUMP_IF_{FALSE,TRUE} from ceval.c to bytecodes.c.
They are no less special than before, but this way
they are discoverable o the copy-and-patch tooling.
During superblock generation, a JUMP_BACKWARD instruction is translated to either a JUMP_TO_TOP micro-op (when the target of the jump is exactly the beginning of the superblock, closing the loop), or a SAVE_IP + EXIT_TRACE pair, when the jump goes elsewhere.
The new JUMP_TO_TOP instruction includes a CHECK_EVAL_BREAKER() call, so a closed loop can still be interrupted.
- Hand-written uops JUMP_IF_{TRUE,FALSE}.
These peek at the top of the stack.
The jump target (in superblock space) is absolute.
- Hand-written translation for POP_JUMP_IF_{TRUE,FALSE},
assuming the jump is unlikely.
Once we implement jump-likelihood profiling,
we can implement the jump-unlikely case (in another PR).
- Tests (including some test cleanup).
- Improvements to len(ex) and ex[i] to expose the whole trace.
When `_PyOptimizer_BackEdge` returns `NULL`, we should restore `next_instr` (and `stack_pointer`). To accomplish this we should jump to `resume_with_error` instead of just `error`.
The problem this causes is subtle -- the only repro I have is in PR gh-106393, at commit d7df54b139bcc47f5ea094bfaa9824f79bc45adc. But the fix is real (as shown later in that PR).
While we're at it, also improve the debug output: the offsets at which traces are identified are now measured in bytes, and always show the start offset. This makes it easier to correlate executor calls with optimizer calls, and either with `dis` output.
<!-- gh-issue-number: gh-104584 -->
* Issue: gh-104584
<!-- /gh-issue-number -->
- Tweak uops debugging output
- Fix the bug from gh-106290
- Rename `SET_IP` to `SAVE_IP` (per https://github.com/faster-cpython/ideas/issues/558)
- Add a `SAVE_IP` uop at the start of the trace (ditto)
- Allow `unbound_local_error`; this gives us uops for `LOAD_FAST_CHECK`, `LOAD_CLOSURE`, and `DELETE_FAST`
- Longer traces
- Support `STORE_FAST_LOAD_FAST`, `STORE_FAST_STORE_FAST`
- Add deps on pycore_uops.h to Makefile(.pre.in)
This produces longer traces (superblocks?).
Also improved debug output (uop names are now printed instead of numeric opcodes). This would be simpler if the numeric opcode values were generated by generate_cases.py, but that's another project.
Refactored some code in generate_cases.py so the essential algorithm for cache effects is only run once. (Deciding which effects are used and what the total cache size is, regardless of what's used.)
This effectively reverts bb578a0, restoring the original DEOPT_IF() macro in ceval_macros.h, and redefining it in the Tier 2 interpreter. We can get rid of the PREDICTED() macros there as well!
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.
* Add table describing possible executable classes for out-of-process debuggers.
* Remove shim code object creation code as it is no longer needed.
* Make lltrace a bit more robust w.r.t. non-standard frames.
For a while now, pending calls only run in the main thread (in the main interpreter). This PR changes things to allow any thread run a pending call, unless the pending call was explicitly added for the main thread to run.
When monitoring LINE events, instrument all instructions that can have a predecessor on a different line.
Then check that the a new line has been hit in the instrumentation code.
This brings the behavior closer to that of 3.11, simplifying implementation and porting of tools.
This speeds up `super()` (by around 85%, for a simple one-level
`super().meth()` microbenchmark) by avoiding allocation of a new
single-use `super()` object on each use.
This is the implementation of PEP683
Motivation:
The PR introduces the ability to immortalize instances in CPython which bypasses reference counting. Tagging objects as immortal allows up to skip certain operations when we know that the object will be around for the entire execution of the runtime.
Note that this by itself will bring a performance regression to the runtime due to the extra reference count checks. However, this brings the ability of having truly immutable objects that are useful in other contexts such as immutable data sharing between sub-interpreters.
* 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
* Rename local variables, names and consts, from the interpeter loop. Will allow non-code objects in frames for better introspection of C builtins and extensions.
* Remove unused dummy variables.
This change is almost entirely moving code around and hiding import state behind internal API. We introduce no changes to behavior, nor to non-internal API. (Since there was already going to be a lot of churn, I took this as an opportunity to re-organize import.c into topically-grouped sections of code.) The motivation is to simplify a number of upcoming changes.
Specific changes:
* move existing import-related code to import.c, wherever possible
* add internal API for interacting with import state (both global and per-interpreter)
* use only API outside of import.c (to limit churn there when changing the location, etc.)
* consolidate the import-related state of PyInterpreterState into a single struct field (this changes layout slightly)
* add macros for import state in import.c (to simplify changing the location)
* group code in import.c into sections
*remove _PyState_AddModule()
https://github.com/python/cpython/issues/101758
* Make sure that the current exception is always normalized.
* Remove redundant type and traceback fields for the current exception.
* Add new API functions: PyErr_GetRaisedException, PyErr_SetRaisedException
* Add new API functions: PyException_GetArgs, PyException_SetArgs
* Adds EXIT_INTERPRETER instruction to exit PyEval_EvalDefault()
* Simplifies RETURN_VALUE, YIELD_VALUE and RETURN_GENERATOR instructions as they no longer need to check for entry frames.
The switch cases (really TARGET(opcode) macros) have been moved from ceval.c to generated_cases.c.h. That file is generated from instruction definitions in bytecodes.c (which impersonates a C file so the C code it contains can be edited without custom support in e.g. VS Code).
The code generator lives in Tools/cases_generator (it has a README.md explaining how it works). The DSL used to describe the instructions is a work in progress, described in https://github.com/faster-cpython/ideas/blob/main/3.12/interpreter_definition.md.
This is surely a work-in-progress. An easy next step could be auto-generating super-instructions.
**IMPORTANT: Merge Conflicts**
If you get a merge conflict for instruction implementations in ceval.c, your best bet is to port your changes to bytecodes.c. That file looks almost the same as the original cases, except instead of `TARGET(NAME)` it uses `inst(NAME)`, and the trailing `DISPATCH()` call is omitted (the code generator adds it automatically).
This reduces confusion between jumps at the bytecode level
(e.g. JUMPTO(), JUMPBY(), and various JUMP_*() opcodes)
and jumps in the C code (which are 'goto' statements).
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).
Make sys.setprofile() and sys.settrace() functions reentrant. They
can no long fail with: RuntimeError("Cannot install a trace function
while another trace function is being installed").
Make _PyEval_SetTrace() and _PyEval_SetProfile() functions reentrant,
rather than detecting and rejecting reentrant calls. Only delete the
reference to function arguments once the new function is fully set,
when a reentrant call is safe. Call also _PySys_Audit() earlier.
The `}` marked with `/* End instructions */` is the end of the switch.
There is another pair of `{}` around the switch, which is vestigial
from ancient times when it was `for (;;) { switch (opcode) { ... } }`.
All `DISPATCH` macro calls should be inside that pair.
Remove the sys.getdxp() function and the Tools/scripts/analyze_dxp.py
script. DXP stands for "dynamic execution pairs". They were related
to DYNAMIC_EXECUTION_PROFILE and DXPAIRS macros which have been
removed in Python 3.11. Python can now be built with "./configure
--enable-pystats" to gather statistics on Python opcodes.
* gh-93503: Add APIs to set profiling and tracing functions in all threads in the C-API
* Use a separate API
* Fix NEWS entry
* Add locks around the loop
* Document ignoring exceptions
* Use the new APIs in the sys module
* Update docs
Move the follow functions and type from frameobject.h to pyframe.h,
so the standard <Python.h> provide frame getter functions:
* PyFrame_Check()
* PyFrame_GetBack()
* PyFrame_GetBuiltins()
* PyFrame_GetGenerator()
* PyFrame_GetGlobals()
* PyFrame_GetLasti()
* PyFrame_GetLocals()
* PyFrame_Type
Remove #include "frameobject.h" from many C files. It's no longer
needed.
This was added for bpo-40514 (gh-84694) to test out a per-interpreter GIL. However, it has since proven unnecessary to keep the experiment in the repo. (It can be done as a branch in a fork like normal.) So here we are removing:
* the configure option
* the macro
* the code enabled by the macro
Fix __lltrace__ debug feature if the stdout encoding is not UTF-8.
If the stdout encoding is not UTF-8, the first call to
lltrace_resume_frame() indirectly sets lltrace to 0 when calling
unicode_check_encoding_errors() which calls
encodings.search_function().
Currently, calling Py_EnterRecursiveCall() and
Py_LeaveRecursiveCall() may use a function call or a static inline
function call, depending if the internal pycore_ceval.h header file
is included or not. Use a different name for the static inline
function to ensure that the static inline function is always used in
Python internals for best performance. Similar approach than
PyThreadState_GET() (function call) and _PyThreadState_GET() (static
inline function).
* Rename _Py_EnterRecursiveCall() to _Py_EnterRecursiveCallTstate()
* Rename _Py_LeaveRecursiveCall() to _Py_LeaveRecursiveCallTstate()
* pycore_ceval.h: Rename Py_EnterRecursiveCall() to
_Py_EnterRecursiveCall() and Py_LeaveRecursiveCall() and
_Py_LeaveRecursiveCall()
* Check the types of PRECALL_METHOD_DESCRIPTOR_FAST_WITH_KEYWORDS
* fix PRECALL_NO_KW_METHOD_DESCRIPTOR_NOARGS as well
* fix PRECALL_NO_KW_METHOD_DESCRIPTOR_O
* fix PRECALL_NO_KW_METHOD_DESCRIPTOR_FAST
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
Macros Py_DECREF, Py_XDECREF, Py_IS_TYPE, _Py_atomic_load_32bit_impl
and _Py_DECREF_SPECIALIZED are redefined as macros
that completely replace the inline functions of the same name.
These three came out in the top four of functions that (in MSVC)
somehow weren't inlined.
Co-authored-by: Victor Stinner <vstinner@python.org>
Co-authored-by: Dennis Sweeney <36520290+sweeneyde@users.noreply.github.com>
Apparently a switch on an 8-bit quantity where all cases are
present generates a more efficient jump (doing only one indexed
memory load instead of two).
So we make opcode and use_tracing uint8_t, and generate a macro
full of extra `case NNN:` lines for all unused opcodes.
See https://github.com/faster-cpython/ideas/issues/321#issuecomment-1103263673
Mark Shannon
Victor Stinner
Brandt Bucher
Guido van Rossum
Irit Katriel
Serhiy Storchaka
Benjamin Peterson
Carl Meyer
Dong-hee Na
Eric Snow
Jelle Zijlstra
Ken Jin
Eddie Elizondo
Nikita Sobolev
Petr Viktorin
Owain Davies
Steve Dower
Eclips4
Matthieu Dartiailh
Kumar Aditya
Skip Montanaro
Dennis Sweeney
Pablo Galindo Salgado
serge-sans-paille
adphrost
Kevin Modzelewski
Itamar Ostricher
Ken Jin
zhanpon
neonene
Ken Jin
Crowthebird