torch/csrc/dynamo/eval_frame.c

#define PY_SSIZE_T_CLEAN
#include <torch/csrc/utils/python_compat.h>
#include <opcode.h>
#include <stdbool.h>

// see https://bugs.python.org/issue35886
#if PY_VERSION_HEX >= 0x03080000
#define Py_BUILD_CORE
#include <internal/pycore_pystate.h>

// These headers were added in 3.11
#if IS_PYTHON_3_11_PLUS
#include <internal/pycore_frame.h>
#define NEED_OPCODE_TABLES // To get _PyOpcode_Deopt
#include <internal/pycore_opcode.h>
#undef NEED_OPCODE_TABLES
#endif

#undef Py_BUILD_CORE
#endif // PY_VERSION_HEX >= 0x03080000

// All the eval APIs change in 3.11 so we need to decide which one to use on the fly
// https://docs.python.org/3/c-api/init.html#c._PyFrameEvalFunction
#if IS_PYTHON_3_11_PLUS
#define THP_EVAL_API_FRAME_OBJECT _PyInterpreterFrame

// The next two functions are taken from
// https://github.com/python/cpython/blob/a7715ccfba5b86ab09f86ec56ac3755c93b46b48/Objects/frameobject.c#L1182
// These are not exported by the CPython binary and thus we have
// to get our own implementation of them.
// As a simple way to reduce the impact of ABI changes on the CPython side, this check forces
// us to manually re-check that the function didn't change on the next major version
#if PY_VERSION_HEX >= 0x030C0000 // 3.12
#error "Please ensure that the functions below still match the CPython implementation for 3.12"
#endif

static int
_PyFrame_OpAlreadyRan(_PyInterpreterFrame *frame, int opcode, int oparg)
{
    // This only works when opcode is a non-quickened form:
    assert(_PyOpcode_Deopt[opcode] == opcode);
    int check_oparg = 0;
    for (_Py_CODEUNIT *instruction = _PyCode_CODE(frame->f_code);
         instruction < frame->prev_instr; instruction++)
    {
        int check_opcode = _PyOpcode_Deopt[_Py_OPCODE(*instruction)];
        check_oparg |= _Py_OPARG(*instruction);
        if (check_opcode == opcode && check_oparg == oparg) {
            return 1;
        }
        if (check_opcode == EXTENDED_ARG) {
            check_oparg <<= 8;
        }
        else {
            check_oparg = 0;
        }
        instruction += _PyOpcode_Caches[check_opcode];
    }
    return 0;
}

int
THP_PyFrame_FastToLocalsWithError(_PyInterpreterFrame *frame) {
    /* Merge fast locals into f->f_locals */
    PyObject *locals;
    PyObject **fast;
    PyCodeObject *co;
    locals = frame->f_locals;
    if (locals == NULL) {
        locals = frame->f_locals = PyDict_New();
        if (locals == NULL)
            return -1;
    }
    co = frame->f_code;
    fast = _PyFrame_GetLocalsArray(frame);
    // COPY_FREE_VARS has no quickened forms, so no need to use _PyOpcode_Deopt
    // here:
    int lasti = _PyInterpreterFrame_LASTI(frame);
    if (lasti < 0 && _Py_OPCODE(_PyCode_CODE(co)[0]) == COPY_FREE_VARS) {
        /* Free vars have not been initialized -- Do that */
        PyCodeObject *co = frame->f_code;
        PyObject *closure = frame->f_func->func_closure;
        int offset = co->co_nlocals + co->co_nplaincellvars;
        for (int i = 0; i < co->co_nfreevars; ++i) {
            PyObject *o = PyTuple_GET_ITEM(closure, i);
            Py_INCREF(o);
            frame->localsplus[offset + i] = o;
        }
        // COPY_FREE_VARS doesn't have inline CACHEs, either:
        frame->prev_instr = _PyCode_CODE(frame->f_code);
    }
    for (int i = 0; i < co->co_nlocalsplus; i++) {
        _PyLocals_Kind kind = _PyLocals_GetKind(co->co_localspluskinds, i);

        /* If the namespace is unoptimized, then one of the
           following cases applies:
           1. It does not contain free variables, because it
              uses import * or is a top-level namespace.
           2. It is a class namespace.
           We don't want to accidentally copy free variables
           into the locals dict used by the class.
        */
        if (kind & CO_FAST_FREE && !(co->co_flags & CO_OPTIMIZED)) {
            continue;
        }

        PyObject *name = PyTuple_GET_ITEM(co->co_localsplusnames, i);
        PyObject *value = fast[i];
        if (frame->stacktop) {
            if (kind & CO_FAST_FREE) {
                // The cell was set by COPY_FREE_VARS.
                assert(value != NULL && PyCell_Check(value));
                value = PyCell_GET(value);
            }
            else if (kind & CO_FAST_CELL) {
                // Note that no *_DEREF ops can happen before MAKE_CELL
                // executes.  So there's no need to duplicate the work
                // that MAKE_CELL would otherwise do later, if it hasn't
                // run yet.
                if (value != NULL) {
                    if (PyCell_Check(value) &&
                            _PyFrame_OpAlreadyRan(frame, MAKE_CELL, i)) {
                        // (likely) MAKE_CELL must have executed already.
                        value = PyCell_GET(value);
                    }
                    // (likely) Otherwise it it is an arg (kind & CO_FAST_LOCAL),
                    // with the initial value set when the frame was created...
                    // (unlikely) ...or it was set to some initial value by
                    // an earlier call to PyFrame_LocalsToFast().
                }
            }
        }
        else {
            assert(value == NULL);
        }
        if (value == NULL) {
            if (PyObject_DelItem(locals, name) != 0) {
                if (PyErr_ExceptionMatches(PyExc_KeyError)) {
                    PyErr_Clear();
                }
                else {
                    return -1;
                }
            }
        }
        else {
            if (PyObject_SetItem(locals, name, value) != 0) {
                return -1;
            }
        }
    }
    return 0;
}

// We need to be able to return the _PyInterpreterFrame to python so create
// a python binding for it

typedef struct THPPyInterpreterFrame {
  PyObject_HEAD
  _PyInterpreterFrame* frame; // Borrowed reference
} THPPyInterpreterFrame;

THPPyInterpreterFrame* THPPyInterpreterFrame_New(_PyInterpreterFrame* frame);

#define DECLARE_PYOBJ_ATTR(name) \
static PyObject* THPPyInterpreterFrame_##name(THPPyInterpreterFrame* self, PyObject* _noargs) { \
  PyObject* res = (PyObject*)self->frame->name; \
  Py_XINCREF(res); \
  return res; \
}

DECLARE_PYOBJ_ATTR(f_func)
DECLARE_PYOBJ_ATTR(f_globals)
DECLARE_PYOBJ_ATTR(f_builtins)
DECLARE_PYOBJ_ATTR(f_locals)
DECLARE_PYOBJ_ATTR(f_code)
DECLARE_PYOBJ_ATTR(frame_obj)

#undef DECLARE_PYOBJ_ATTR

static THPPyInterpreterFrame* THPPyInterpreterFrame_previous(THPPyInterpreterFrame* self, PyObject* _noargs) {
  THPPyInterpreterFrame* res = THPPyInterpreterFrame_New(self->frame->previous);
  return res;
}

// This is not a true attribute of the class but we do access it in python and it is hard to implement
// on the python side, so do it here:
static PyObject* THPPyInterpreterFrame_f_lasti(THPPyInterpreterFrame* self, PyObject* _noargs) {
  return PyLong_FromLong(_PyInterpreterFrame_LASTI(self->frame));
}

// NOLINTNEXTLINE(cppcoreguidelines-avoid-c-arrays,cppcoreguidelines-avoid-non-const-global-variables,modernize-avoid-c-arrays)
static struct PyGetSetDef THPDevice_properties[] = {
    {"f_func", (getter)THPPyInterpreterFrame_f_func, NULL, NULL, NULL},
    {"f_globals", (getter)THPPyInterpreterFrame_f_globals, NULL, NULL, NULL},
    {"f_builtins", (getter)THPPyInterpreterFrame_f_builtins, NULL, NULL, NULL},
    {"f_locals", (getter)THPPyInterpreterFrame_f_locals, NULL, NULL, NULL},
    {"f_code", (getter)THPPyInterpreterFrame_f_code, NULL, NULL, NULL},
    {"frame_obj", (getter)THPPyInterpreterFrame_frame_obj, NULL, NULL, NULL},
    {"previous", (getter)THPPyInterpreterFrame_previous, NULL, NULL, NULL},
    {"f_lasti", (getter)THPPyInterpreterFrame_f_lasti, NULL, NULL, NULL},
    {NULL}};

PyTypeObject THPPyInterpreterFrameType = {
    PyVarObject_HEAD_INIT(NULL, 0) "torch._C.dynamo.eval_frame._PyInterpreterFrame", /* tp_name */
    sizeof(THPPyInterpreterFrame), /* tp_basicsize */
    0, /* tp_itemsize */
    NULL, /* tp_dealloc */
    0, /* tp_vectorcall_offset */
    NULL, /* tp_getattr */
    NULL, /* tp_setattr */
    NULL, /* tp_reserved */
    NULL, /* tp_repr */
    NULL, /* tp_as_number */
    NULL, /* tp_as_sequence */
    NULL, /* tp_as_mapping */
    NULL, /* tp_hash  */
    NULL, /* tp_call */
    NULL, /* tp_str */
    NULL, /* tp_getattro */
    NULL, /* tp_setattro */
    NULL, /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT, /* tp_flags */
    NULL, /* tp_doc */
    NULL, /* tp_traverse */
    NULL, /* tp_clear */
    NULL, /* tp_richcompare */
    0, /* tp_weaklistoffset */
    NULL, /* tp_iter */
    NULL, /* tp_iternext */
    NULL, /* tp_methods */
    NULL, /* tp_members */
    THPDevice_properties, /* tp_getset */
    NULL, /* tp_base */
    NULL, /* tp_dict */
    NULL, /* tp_descr_get */
    NULL, /* tp_descr_set */
    0, /* tp_dictoffset */
    NULL, /* tp_init */
    NULL, /* tp_alloc */
    NULL, /* tp_new */
};


THPPyInterpreterFrame* THPPyInterpreterFrame_New(_PyInterpreterFrame* frame) {
  PyTypeObject* type = (PyTypeObject*)&THPPyInterpreterFrameType;
  THPPyInterpreterFrame* self = (THPPyInterpreterFrame*)type->tp_alloc(type, 0);
  if (!self)
    return NULL;
  self->frame = frame;
  return self;
}


#else
#define THP_EVAL_API_FRAME_OBJECT PyFrameObject

#define THP_PyFrame_FastToLocalsWithError PyFrame_FastToLocalsWithError
#endif

#ifdef _WIN32
#define unlikely(x) (x)
#else
#define unlikely(x) __builtin_expect((x), 0)
#endif

#define NULL_CHECK(val)                                         \
  if (unlikely((val) == NULL)) {                                \
    fprintf(stderr, "NULL ERROR: %s:%d\n", __FILE__, __LINE__); \
    PyErr_Print();                                              \
    abort();                                                    \
  } else {                                                      \
  }

#define CHECK(cond)                                                     \
  if (unlikely(!(cond))) {                                              \
    fprintf(stderr, "DEBUG CHECK FAILED: %s:%d\n", __FILE__, __LINE__); \
    abort();                                                            \
  } else {                                                              \
  }

#ifdef TORCHDYNAMO_DEBUG

#define DEBUG_CHECK(cond) CHECK(cond)
#define DEBUG_NULL_CHECK(val) NULL_CHECK(val)
#define DEBUG_TRACE(msg, ...) \
  fprintf(stderr, "TRACE[%s:%d] " msg "\n", __func__, __LINE__, __VA_ARGS__)
#define DEBUG_TRACE0(msg) \
  fprintf(stderr, "TRACE[%s:%d] " msg "\n", __func__, __LINE__)

#else

#define DEBUG_CHECK(cond)
#define DEBUG_NULL_CHECK(val)
#define DEBUG_TRACE(msg, ...)
#define DEBUG_TRACE0(msg)

#endif

// Flag to just run a frame normally
#define SKIP_CODE ((void*)0x1)

static PyObject* noargs = NULL; /* cached empty tuple */
static PyObject* dotzerokey = NULL; /* ".0" */
static PyObject* guard_fail_hook = NULL;
static PyObject* guard_error_hook = NULL;
static PyObject* profiler_start_hook = NULL;
static PyObject* profiler_end_hook = NULL;
static PyObject* guard_profiler_name_str = NULL; /* cached py str */

size_t extra_index = -1;

static Py_tss_t eval_frame_callback_key = Py_tss_NEEDS_INIT;

inline static PyObject* eval_frame_callback_get(void) {
  void* result = PyThread_tss_get(&eval_frame_callback_key);
  if (unlikely(result == NULL)) {
    Py_RETURN_NONE;
  } else {
    return (PyObject*)result;
  }
}

inline static void eval_frame_callback_set(PyObject* obj) {
  PyThread_tss_set(&eval_frame_callback_key, obj);
}

static void ignored(void* obj) {}
static PyObject* _custom_eval_frame_shim(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    int throw_flag);
static PyObject* _custom_eval_frame(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    int throw_flag,
    PyObject* callback);
static PyObject *(*previous_eval_frame)(PyThreadState *tstate,
                                        PyFrameObject *frame, int throw_flag) = NULL;

#if PY_VERSION_HEX >= 0x03090000
static PyObject* custom_eval_frame_shim(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    int throw_flag) {
  return _custom_eval_frame_shim(tstate, frame, throw_flag);
}
#else
static PyObject* custom_eval_frame_shim(THP_EVAL_API_FRAME_OBJECT* frame, int throw_flag) {
  PyThreadState* tstate = PyThreadState_GET();
  return _custom_eval_frame_shim(tstate, frame, throw_flag);
}
#endif

inline static PyObject* eval_frame_default(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    int throw_flag) {
#if PY_VERSION_HEX >= 0x03090000
  if (tstate == NULL) {
    tstate = PyThreadState_GET();
  }
  if (previous_eval_frame) {
    return previous_eval_frame(tstate, frame, throw_flag);
  }
  else {
    return _PyEval_EvalFrameDefault(tstate, frame, throw_flag);
  }
#else
  return _PyEval_EvalFrameDefault(frame, throw_flag);
#endif
}

inline static void enable_eval_frame_shim(PyThreadState* tstate) {
#if PY_VERSION_HEX >= 0x03090000
  if (_PyInterpreterState_GetEvalFrameFunc(tstate->interp) !=
      &custom_eval_frame_shim) {
    DEBUG_CHECK(previous_eval_frame == NULL);
    previous_eval_frame = _PyInterpreterState_GetEvalFrameFunc(tstate->interp);
    _PyInterpreterState_SetEvalFrameFunc(tstate->interp,
                                         &custom_eval_frame_shim);
  }
#else
  if (tstate->interp->eval_frame != &custom_eval_frame_shim) {
    // First call
    tstate->interp->eval_frame = &custom_eval_frame_shim;
  }
#endif
}

inline static void enable_eval_frame_default(PyThreadState* tstate) {
#if PY_VERSION_HEX >= 0x03090000
  if (_PyInterpreterState_GetEvalFrameFunc(tstate->interp) !=
      previous_eval_frame) {
    DEBUG_CHECK(previous_eval_frame != NULL);
    _PyInterpreterState_SetEvalFrameFunc(tstate->interp,
                                         previous_eval_frame);
    previous_eval_frame = NULL;
  }
#else
  if (tstate->interp->eval_frame != &_PyEval_EvalFrameDefault) {
    // First call
    tstate->interp->eval_frame = &_PyEval_EvalFrameDefault;
  }
#endif
}

static inline PyObject* call_callback(
    PyObject* callable,
    THP_EVAL_API_FRAME_OBJECT* _frame,
    long cache_len) {

#if IS_PYTHON_3_11_PLUS
  THPPyInterpreterFrame* frame = THPPyInterpreterFrame_New(_frame);
#else
  PyFrameObject* frame = _frame;
#endif
  PyObject* args = Py_BuildValue("(Ol)", frame, cache_len);
  if (args == NULL) {
    return NULL;
  }
  PyObject* result = PyObject_CallObject(callable, args);
  Py_DECREF(args);
  return result;
}

typedef struct cache_entry {
  // check the guards: lambda: <locals of user function>: bool
  PyObject* check_fn;
  // modified user bytecode (protected by check_fn's guards)
  PyCodeObject* code;
  // on a cache miss, linked list of next thing to try
  struct cache_entry* next;
} CacheEntry;

static CacheEntry* create_cache_entry(
    CacheEntry* next,
    PyObject* guarded_code) {
  CacheEntry* e = (CacheEntry*)malloc(sizeof(CacheEntry));
  DEBUG_NULL_CHECK(e);
  e->check_fn = PyObject_GetAttrString(guarded_code, "check_fn");
  NULL_CHECK(e->check_fn);
  e->code = (PyCodeObject*)PyObject_GetAttrString(guarded_code, "code");
  NULL_CHECK(e->code);
  e->next = next;
  return e;
}

static void destroy_cache_entry(CacheEntry* e) {
  if (e == NULL || e == SKIP_CODE) {
    return;
  }
  Py_XDECREF(e->check_fn);
  Py_XDECREF(e->code);
  destroy_cache_entry(e->next);
  free(e);
}

inline static CacheEntry* get_extra(PyCodeObject* code) {
  CacheEntry* extra = NULL;
  _PyCode_GetExtra((PyObject*)code, extra_index, (void*)&extra);
  return extra;
}

inline static void set_extra(PyCodeObject* code, CacheEntry* extra) {
  // TODO(jansel): would it be faster to bypass this?
  _PyCode_SetExtra((PyObject*)code, extra_index, extra);
}

inline static const char* name(THP_EVAL_API_FRAME_OBJECT* frame) {
  DEBUG_CHECK(PyUnicode_Check(frame->f_code->co_name));
  return PyUnicode_AsUTF8(frame->f_code->co_name);
}

static PyObject* call_guard_fail_hook(
    PyObject* hook,
    CacheEntry* e,
    size_t index,
    PyObject* f_locals) {
  // call debugging logic when a guard fails
  PyObject* args = PyTuple_Pack(
      5,
      e->check_fn,
      e->code,
      f_locals,
      PyLong_FromSize_t(index),
      (e->next == NULL ? Py_True : Py_False));
  if (args == NULL) return NULL;
  PyObject* result = PyObject_CallObject(hook, args);
  Py_DECREF(args);
  return result;
}

static PyObject* call_profiler_start_hook(PyObject* name_str) {
  if (profiler_start_hook == NULL) return NULL;
  if (name_str == NULL) return NULL;
  PyObject* args = PyTuple_Pack(1, name_str);
  if (args == NULL) return NULL;
  PyObject* result = PyObject_CallObject(profiler_start_hook, args);
  Py_DECREF(args);
  return result;
}

static void call_profiler_end_hook(PyObject* record) {
  // 'record' obj is the return value of calling _start_hook()
  if (profiler_end_hook == NULL) return;
  if (record == NULL) return;
  PyObject* args = PyTuple_Pack(1, record);
  if (args == NULL) return;
  PyObject* result = PyObject_CallObject(profiler_end_hook, args);
  Py_XDECREF(result);
  Py_DECREF(args);
}

// Return value: borrowed reference
// Is either Py_None or a PyCodeObject
static PyObject* lookup(CacheEntry* e, THP_EVAL_API_FRAME_OBJECT *frame, CacheEntry* prev, size_t index) {
  if (e == NULL) {
    // NB: intentionally not using Py_RETURN_NONE, to return borrowed ref
    return Py_None;
  }
  PyObject *f_locals = frame->f_locals;
  PyObject* dotzero = PyDict_GetItem(f_locals, dotzerokey);
  PyObject* valid = NULL;
  if (unlikely(dotzero != NULL)) {
    // .0 is a special variable name used for implicit args
    PyObject* args = PyTuple_Pack(1, dotzero);
    if (args == NULL) return NULL;
    valid = PyObject_Call(e->check_fn, args, f_locals);
    Py_DECREF(args);
  } else {
    valid = PyObject_Call(e->check_fn, noargs, f_locals);
  }
  if (unlikely(valid == NULL)) {
    if (guard_error_hook != NULL) {
      PyObject *type, *value, *traceback;
      PyErr_Fetch(&type, &value, &traceback);
      PyObject* r = call_guard_fail_hook(guard_error_hook, e, index, f_locals);
      if (r == NULL) {
        return NULL;
      }
      Py_DECREF(r);
      PyErr_Restore(type, value, traceback);
    }
    return NULL;
  }
  Py_DECREF(valid);
  if (valid == Py_True) {
    // Keep the head as the most recently used cache entry.
    // If the hit cache entry is not the head of the linked list,
    // move it to the head
    if (prev != NULL) {
        CacheEntry* extra = get_extra(frame->f_code);
        prev->next = e->next;
        e->next = extra;
        set_extra(frame->f_code, e);
    }
    return (PyObject*)e->code;
  }
  if (unlikely(guard_fail_hook != NULL)) {
    PyObject* r = call_guard_fail_hook(guard_fail_hook, e, index, f_locals);
    if (r == NULL) {
      return NULL;
    }
    Py_DECREF(r);
  }
  return lookup(e->next, frame, e, index + 1);
}

static long cache_size(CacheEntry* e) {
  if (e == NULL) {
    return 0;
  }
  return 1 + cache_size(e->next);
}

inline static PyObject* eval_custom_code(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    PyCodeObject* code,
    int throw_flag) {
  Py_ssize_t ncells = 0;
  Py_ssize_t nfrees = 0;
  Py_ssize_t nlocals_new = code->co_nlocals;
  Py_ssize_t nlocals_old = frame->f_code->co_nlocals;

  ncells = PyCode_GetNCellvars(code);
  nfrees = PyCode_GetNFreevars(code);

  DEBUG_NULL_CHECK(tstate);
  DEBUG_NULL_CHECK(frame);
  DEBUG_NULL_CHECK(code);
  #if IS_PYTHON_3_11_PLUS
  DEBUG_CHECK(ncells == frame->f_code->co_ncellvars);
  DEBUG_CHECK(nfrees == frame->f_code->co_nfreevars);
  #else
  DEBUG_CHECK(ncells == PyTuple_GET_SIZE(frame->f_code->co_cellvars));
  DEBUG_CHECK(nfrees == PyTuple_GET_SIZE(frame->f_code->co_freevars));
  #endif
  DEBUG_CHECK(nlocals_new >= nlocals_old);

  PyFrameObject* shadow_obj = PyFrame_New(tstate, code, frame->f_globals, NULL);
  #if IS_PYTHON_3_11_PLUS
  THP_EVAL_API_FRAME_OBJECT* shadow = shadow_obj->f_frame;
  Py_XINCREF(frame->f_func->func_closure);
  shadow->f_func->func_closure = frame->f_func->func_closure;
  #else
  THP_EVAL_API_FRAME_OBJECT* shadow = shadow_obj;
  #endif
  if (shadow == NULL) {
    Py_DECREF(shadow_obj);
    return NULL;
  }

  #if IS_PYTHON_3_11_PLUS
  PyObject** fastlocals_old = frame->localsplus;
  PyObject** fastlocals_new = shadow->localsplus;

  // copy from old fastlocals to new fastlocals:
  // for i, name in enumerate(localsplusnames_new):
  //   name_to_idx[name] = i
  // for i, name in enumerate(localsplusnames_old):
  //   fastlocals_new[name_to_idx[name]] = fastlocals_old[i]
  PyObject* name_to_idx = PyDict_New();
  if (name_to_idx == NULL) {
    DEBUG_TRACE0("unable to create localsplus name dict");
    Py_DECREF(shadow_obj);
    return NULL;
  }

  for (Py_ssize_t i = 0; i < code->co_nlocalsplus; i++) {
    PyObject *name = PyTuple_GET_ITEM(code->co_localsplusnames, i);
    PyObject *idx = PyLong_FromSsize_t(i);
    if (name == NULL || idx == NULL || PyDict_SetItem(name_to_idx, name, idx) != 0) {
      Py_DECREF(shadow_obj);
      Py_DECREF(name_to_idx);
      return NULL;
    }
  }

  for (Py_ssize_t i = 0; i < frame->f_code->co_nlocalsplus; i++) {
    PyObject *name = PyTuple_GET_ITEM(frame->f_code->co_localsplusnames, i);
    PyObject *idx = PyDict_GetItem(name_to_idx, name);
    Py_ssize_t new_i = PyLong_AsSsize_t(idx);
    if (name == NULL || idx == NULL || (new_i == (Py_ssize_t)-1 && PyErr_Occurred() != NULL)) {
      Py_DECREF(shadow_obj);
      Py_DECREF(name_to_idx);
      return NULL;
    }
    Py_XINCREF(fastlocals_old[i]);
    fastlocals_new[new_i] = fastlocals_old[i];
  }

  Py_DECREF(name_to_idx);
  #else
  PyObject** fastlocals_old = frame->f_localsplus;
  PyObject** fastlocals_new = shadow->f_localsplus;

  for (Py_ssize_t i = 0; i < nlocals_old; i++) {
    Py_XINCREF(fastlocals_old[i]);
    fastlocals_new[i] = fastlocals_old[i];
  }

  for (Py_ssize_t i = 0; i < ncells + nfrees; i++) {
    Py_XINCREF(fastlocals_old[nlocals_old + i]);
    fastlocals_new[nlocals_new + i] = fastlocals_old[nlocals_old + i];
  }
  #endif

  PyObject* result = eval_frame_default(tstate, shadow, throw_flag);
  Py_DECREF(shadow_obj);
  return result;
}

static PyObject* _custom_eval_frame_shim(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    int throw_flag) {
  // Shims logic into one of three states. Can probably be refactored into a
  // single func, later:
  //  - None: disables TorchDynamo
  //  - False: run-only mode (reuse existing compiles)
  //  - Python callable(): enables TorchDynamo
  PyObject* callback = eval_frame_callback_get();

  if (callback == Py_None) {
    return eval_frame_default(tstate, frame, throw_flag);
  }

  return _custom_eval_frame(tstate, frame, throw_flag, callback);
}

static PyObject* _custom_eval_frame(
    PyThreadState* tstate,
    THP_EVAL_API_FRAME_OBJECT* frame,
    int throw_flag,
    PyObject* callback) {
  #if IS_PYTHON_3_11_PLUS
  DEBUG_TRACE(
      "begin %s %s %i %i",
      name(frame),
      PyUnicode_AsUTF8(frame->f_code->co_filename),
      frame->f_code->co_firstlineno,
      _PyInterpreterFrame_LASTI(frame));
  #else
  DEBUG_TRACE(
      "begin %s %s %i %i %i",
      name(frame),
      PyUnicode_AsUTF8(frame->f_code->co_filename),
      frame->f_lineno,
      frame->f_lasti,
      frame->f_iblock);
  #endif

  if (throw_flag) {
    // When unwinding generators, eval frame is called with throw_flag ==
    // true.  Frame evaluation is supposed to continue unwinding by propagating
    // the exception.  Dynamo doesn't really know how to do this, nor does it
    // really want to do this, because there's unlikely any code to capture
    // (you're going to immediately quit out of the frame, perhaps running
    // some unwinding logic along the way).  So we just run the default
    // handler in this case.
    //
    // NB: A previous version of this patch returned NULL.  This is wrong,
    // because returning NULL is *different* from unwinding an exception.
    // In particular, you will not execute things like context manager
    // __exit__ if you just return NULL.
    //
    // NB: It's /conceivable/ that you might want to actually still call the
    // Dynamo callback when throw_flag == TRUE, to give Dynamo a chance to
    // do any stack unwinding code.  But this is not really useful because
    // (1) Dynamo doesn't actually know how to do stack unwinding, so it would
    // immediately skip the frame, and (2) even if it did, this would only
    // be profitable if there was tensor code in the unwinding code.  Seems
    // unlikely.
    DEBUG_TRACE("throw %s", name(frame));
    return eval_frame_default(tstate, frame, throw_flag);
  }

  CacheEntry* extra = get_extra(frame->f_code);
  if (extra == SKIP_CODE || (callback == Py_False && extra == NULL)) {
    DEBUG_TRACE("skip %s", name(frame));
    return eval_frame_default(tstate, frame, throw_flag);
  }

  // TODO(jansel): investigate directly using the "fast" representation
  // TODO(alband): This is WRONG for python3.11+ we pass in a _PyInterpreterFrame
  // even though we should pass a PyFrameObject.
  if (THP_PyFrame_FastToLocalsWithError(frame) < 0) {
    DEBUG_TRACE("error %s", name(frame));
    return NULL;
  }

  // A callback of Py_False indicates "run only" mode, the cache is checked, but
  // we never compile.
  if (callback == Py_False) {
    DEBUG_TRACE("In run only mode %s", name(frame));
    PyObject* hook_record = call_profiler_start_hook(guard_profiler_name_str);
    PyObject* maybe_cached_code = lookup(extra, frame, NULL, 0);
    call_profiler_end_hook(hook_record);
    Py_XDECREF(hook_record);

    if (maybe_cached_code == NULL) {
      // guard eval failed, keep propagating
      return NULL;
    } else if (maybe_cached_code == Py_None) {
      DEBUG_TRACE("cache miss %s", name(frame));
      return eval_frame_default(tstate, frame, throw_flag);
    }
    PyCodeObject* cached_code = (PyCodeObject*)maybe_cached_code;
    // used cached version
    DEBUG_TRACE("cache hit %s", name(frame));
    return eval_custom_code(tstate, frame, cached_code, throw_flag);
  }
  DEBUG_CHECK(PyDict_CheckExact(frame->f_locals));
  DEBUG_CHECK(PyDict_CheckExact(frame->f_globals));
  DEBUG_CHECK(PyDict_CheckExact(frame->f_builtins));

  // We don't run the current custom_eval_frame behavior for guards.
  // So we temporarily set the callback to Py_None to drive the correct behavior
  // in the shim.
  eval_frame_callback_set(Py_None);

  PyObject* hook_record = call_profiler_start_hook(guard_profiler_name_str);
  PyObject* maybe_cached_code = lookup(extra, frame, NULL, 0);
  call_profiler_end_hook(hook_record);
  Py_XDECREF(hook_record);
  if (maybe_cached_code == NULL) {
    // Python error
    return NULL;
  } else if (maybe_cached_code != Py_None) {
    PyCodeObject* cached_code = (PyCodeObject*)maybe_cached_code;
    // used cached version
    DEBUG_TRACE("cache hit %s", name(frame));
    // Re-enable custom behavior
    eval_frame_callback_set(callback);
    return eval_custom_code(tstate, frame, cached_code, throw_flag);
  }
  // cache miss

  // TODO(alband): This is WRONG for python3.11+ we pass in a _PyInterpreterFrame
  // that gets re-interpreted as a PyObject (which it is NOT!)
  PyObject* result =
      call_callback(callback, frame, cache_size(extra));
  if (result == NULL) {
    // internal exception, returning here will leak the exception into user code
    // this is useful for debugging -- but we dont want it to happen outside of
    // testing
    // NB: we intentionally DO NOT re-enable custom behavior to prevent
    // cascading failure from internal exceptions.  The upshot is if
    // Dynamo barfs, that's it for Dynamo, even if you catch the exception
    // inside the torch.compile block we won't try to Dynamo anything else.
    return NULL;
  } else if (result != Py_None) {
    DEBUG_TRACE("create cache %s", name(frame));
    extra = create_cache_entry(extra, result);
    Py_DECREF(result);
    set_extra(frame->f_code, extra);
    // Re-enable custom behavior
    eval_frame_callback_set(callback);
    return eval_custom_code(tstate, frame, extra->code, throw_flag);
  } else {
    DEBUG_TRACE("create skip %s", name(frame));
    Py_DECREF(result);
    destroy_cache_entry(extra);
    set_extra(frame->f_code, SKIP_CODE);
    // Re-enable custom behavior
    eval_frame_callback_set(callback);
    return eval_frame_default(tstate, frame, throw_flag);
  }
}

static int active_dynamo_threads = 0;

static PyObject* increment_working_threads(PyThreadState* tstate) {
  active_dynamo_threads = active_dynamo_threads + 1;
  if (active_dynamo_threads > 0) {
    enable_eval_frame_shim(tstate);
  }
  Py_RETURN_NONE;
}

static PyObject* decrement_working_threads(PyThreadState* tstate) {
  if (active_dynamo_threads > 0) {
    active_dynamo_threads = active_dynamo_threads - 1;
    if (active_dynamo_threads == 0) {
      enable_eval_frame_default(tstate);
    }
  }
  Py_RETURN_NONE;
}

static PyObject* set_eval_frame(PyObject* new_callback, PyThreadState* tstate) {
  // Change the eval frame callback and return the old one
  //  - None: disables TorchDynamo
  //  - False: run-only mode (reuse existing compiles)
  //  - Python callable(): enables TorchDynamo
  PyObject* old_callback = eval_frame_callback_get();

  // owned by caller
  Py_INCREF(old_callback);

  if (old_callback != Py_None && new_callback == Py_None) {
    decrement_working_threads(tstate);
  } else if (old_callback == Py_None && new_callback != Py_None) {
    increment_working_threads(tstate);
  }

  Py_INCREF(new_callback);
  Py_DECREF(old_callback);

  // Set thread local callback. This will drive behavior of our shim, if/when it
  // is installed.
  eval_frame_callback_set(new_callback);

  return old_callback;
}

static PyObject* set_eval_frame_py(PyObject* dummy, PyObject* args) {
  PyObject* callback = NULL;
  if (!PyArg_ParseTuple(args, "O:callback", &callback)) {
    DEBUG_TRACE0("arg error");
    return NULL;
  }
  if (callback != Py_None && callback != Py_False &&
      !PyCallable_Check(callback)) {
    DEBUG_TRACE0("arg error");
    PyErr_SetString(PyExc_TypeError, "expected a callable");
    return NULL;
  }
  DEBUG_TRACE(
      "python enabled=%d and is run_only=%d",
      callback != Py_None,
      callback == Py_False);
  return set_eval_frame(callback, PyThreadState_GET());
}

static PyObject* reset_code(PyObject* dummy, PyObject* args) {
  PyObject* code = NULL;
  if (!PyArg_ParseTuple(args, "O:code", &code)) {
    DEBUG_TRACE0("arg error");
    return NULL;
  }
  if (!PyCode_Check(code)) {
    DEBUG_TRACE0("arg error");
    PyErr_SetString(PyExc_TypeError, "expected a code object");
    return NULL;
  }

  destroy_cache_entry(get_extra((PyCodeObject*)code));
  set_extra((PyCodeObject*)code, NULL);
  Py_RETURN_NONE;
}

static PyObject* unsupported(PyObject* dummy, PyObject* args) {
  // a dummy C function used in testing
  PyObject* obj1 = NULL;
  PyObject* obj2 = NULL;
  if (!PyArg_ParseTuple(args, "OO", &obj1, &obj2)) {
    return NULL;
  }
  Py_INCREF(obj2);
  return obj2;
}

static PyObject* skip_code(PyObject* dummy, PyObject* args) {
  PyObject* obj = NULL;
  if (!PyArg_ParseTuple(args, "O", &obj)) {
    return NULL;
  }
  if (!PyCode_Check(obj)) {
    PyErr_SetString(PyExc_TypeError, "expected a code object");
    return NULL;
  }
  set_extra((PyCodeObject*)obj, SKIP_CODE);
  Py_RETURN_NONE;
}

static PyObject* set_guard_fail_hook(PyObject* dummy, PyObject* args) {
  PyObject* obj = NULL;
  if (!PyArg_ParseTuple(args, "O", &obj)) {
    return NULL;
  }
  Py_XDECREF(guard_fail_hook);
  if (obj == Py_None) {
    guard_fail_hook = NULL;
  } else {
    guard_fail_hook = obj;
    Py_INCREF(guard_fail_hook);
  }
  Py_RETURN_NONE;
}

static PyObject* set_guard_error_hook(PyObject* dummy, PyObject* args) {
  PyObject* obj = NULL;
  if (!PyArg_ParseTuple(args, "O", &obj)) {
    return NULL;
  }
  Py_XDECREF(guard_error_hook);
  if (obj == Py_None) {
    guard_error_hook = NULL;
  } else {
    guard_error_hook = obj;
    Py_INCREF(guard_error_hook);
  }
  Py_RETURN_NONE;
}

static PyObject* clear_profiler_hooks(PyObject* dummy, PyObject* args) {
  Py_XDECREF(profiler_start_hook);
  profiler_start_hook = NULL;
  Py_XDECREF(profiler_end_hook);
  profiler_end_hook = NULL;
  Py_XDECREF(guard_profiler_name_str);
  guard_profiler_name_str = NULL;
  Py_RETURN_NONE;
}

static PyObject* set_profiler_hooks(PyObject* dummy, PyObject* args) {
  PyObject* start = NULL;
  PyObject* end = NULL;
  if (!PyArg_ParseTuple(args, "OO", &start, &end)) {
    return NULL;
  }
  Py_XDECREF(profiler_start_hook);
  Py_XDECREF(profiler_end_hook);
  if (start == Py_None || end == Py_None) {
    clear_profiler_hooks(NULL, NULL);
  } else {
    profiler_start_hook = start;
    profiler_end_hook = end;
    Py_INCREF(profiler_start_hook);
    Py_INCREF(profiler_end_hook);
  }
  Py_XDECREF(guard_profiler_name_str);
  guard_profiler_name_str = Py_BuildValue("s", "TorchDynamo Cache Lookup");
  Py_RETURN_NONE;
}

static PyMethodDef _methods[] = {
    {"set_eval_frame", set_eval_frame_py, METH_VARARGS, NULL},
    {"reset_code", reset_code, METH_VARARGS, NULL},
    {"unsupported", unsupported, METH_VARARGS, NULL},
    {"skip_code", skip_code, METH_VARARGS, NULL},
    {"set_guard_fail_hook", set_guard_fail_hook, METH_VARARGS, NULL},
    {"set_guard_error_hook", set_guard_error_hook, METH_VARARGS, NULL},
    {"set_profiler_hooks", set_profiler_hooks, METH_VARARGS, NULL},
    {"clear_profiler_hooks", clear_profiler_hooks, METH_VARARGS, NULL},
    {NULL, NULL, 0, NULL}};

static struct PyModuleDef _module = {
    PyModuleDef_HEAD_INIT,
    "torch._C._dynamo.eval_frame",
    "Module containing hooks to override eval_frame",
    -1,
    _methods};

PyObject* torch_c_dynamo_eval_frame_init(void) {
  extra_index = _PyEval_RequestCodeExtraIndex(ignored);

  int result = PyThread_tss_create(&eval_frame_callback_key);
  CHECK(result == 0);

  Py_INCREF(Py_None);
  eval_frame_callback_set(Py_None);

  noargs = PyTuple_New(0);
  dotzerokey = PyUnicode_InternFromString(".0");
  PyObject* module = PyModule_Create(&_module);

#if IS_PYTHON_3_11_PLUS
  if (PyType_Ready(&THPPyInterpreterFrameType) < 0) {
    return NULL;
  }
  Py_INCREF(&THPPyInterpreterFrameType);
  if (PyModule_AddObject(module, "_PyInterpreterFrame", (PyObject*)&THPPyInterpreterFrameType) != 0) {
    return NULL;
  }
#endif

  return module;
}