mirror of
https://github.com/python/cpython.git
synced 2025-08-03 16:39:00 +00:00
gh-117045: Add code object to function version cache (#117028)
Changes to the function version cache: - In addition to the function object, also store the code object, and allow the latter to be retrieved even if the function has been evicted. - Stop assigning new function versions after a critical attribute (e.g. `__code__`) has been modified; the version is permanently reset to zero in this case. - Changes to `__annotations__` are no longer considered critical. (This fixes gh-109998.) Changes to the Tier 2 optimization machinery: - If we cannot map a function version to a function, but it is still mapped to a code object, we continue projecting the trace. The operand of the `_PUSH_FRAME` and `_POP_FRAME` opcodes can be either NULL, a function object, or a code object with the lowest bit set. This allows us to trace through code that calls an ephemeral function, i.e., a function that may not be alive when we are constructing the executor, e.g. a generator expression or certain nested functions. We will lose globals removal inside such functions, but we can still do other peephole operations (and even possibly [call inlining](https://github.com/python/cpython/pull/116290), if we decide to do it), which only need the code object. As before, if we cannot retrieve the code object from the cache, we stop projecting.
This commit is contained in:
Guido van Rossum
GitHub
parent
c85d84166a
commit
570a82d46a
8 changed files with 208 additions and 95 deletions
|
|
@ -211,7 +211,7 @@ _PyOptimizer_Optimize(
|
|||
_PyInterpreterFrame *frame, _Py_CODEUNIT *start,
|
||||
PyObject **stack_pointer, _PyExecutorObject **executor_ptr)
|
||||
{
|
||||
PyCodeObject *code = (PyCodeObject *)frame->f_executable;
|
||||
PyCodeObject *code = _PyFrame_GetCode(frame);
|
||||
assert(PyCode_Check(code));
|
||||
PyInterpreterState *interp = _PyInterpreterState_GET();
|
||||
if (!has_space_for_executor(code, start)) {
|
||||
|
|
@ -479,8 +479,9 @@ BRANCH_TO_GUARD[4][2] = {
|
|||
ADD_TO_TRACE(_EXIT_TRACE, 0, 0, 0); \
|
||||
goto done; \
|
||||
} \
|
||||
assert(func->func_code == (PyObject *)code); \
|
||||
assert(func == NULL || func->func_code == (PyObject *)code); \
|
||||
trace_stack[trace_stack_depth].func = func; \
|
||||
trace_stack[trace_stack_depth].code = code; \
|
||||
trace_stack[trace_stack_depth].instr = instr; \
|
||||
trace_stack_depth++;
|
||||
#define TRACE_STACK_POP() \
|
||||
|
|
@ -489,7 +490,8 @@ BRANCH_TO_GUARD[4][2] = {
|
|||
} \
|
||||
trace_stack_depth--; \
|
||||
func = trace_stack[trace_stack_depth].func; \
|
||||
code = (PyCodeObject *)trace_stack[trace_stack_depth].func->func_code; \
|
||||
code = trace_stack[trace_stack_depth].code; \
|
||||
assert(func == NULL || func->func_code == (PyObject *)code); \
|
||||
instr = trace_stack[trace_stack_depth].instr;
|
||||
|
||||
/* Returns 1 on success,
|
||||
|
|
@ -505,7 +507,7 @@ translate_bytecode_to_trace(
|
|||
_PyBloomFilter *dependencies)
|
||||
{
|
||||
bool progress_needed = true;
|
||||
PyCodeObject *code = (PyCodeObject *)frame->f_executable;
|
||||
PyCodeObject *code = _PyFrame_GetCode(frame);
|
||||
PyFunctionObject *func = (PyFunctionObject *)frame->f_funcobj;
|
||||
assert(PyFunction_Check(func));
|
||||
PyCodeObject *initial_code = code;
|
||||
|
|
@ -515,6 +517,7 @@ translate_bytecode_to_trace(
|
|||
int max_length = buffer_size;
|
||||
struct {
|
||||
PyFunctionObject *func;
|
||||
PyCodeObject *code;
|
||||
_Py_CODEUNIT *instr;
|
||||
} trace_stack[TRACE_STACK_SIZE];
|
||||
int trace_stack_depth = 0;
|
||||
|
|
@ -719,9 +722,19 @@ top: // Jump here after _PUSH_FRAME or likely branches
|
|||
|
||||
if (uop == _POP_FRAME) {
|
||||
TRACE_STACK_POP();
|
||||
/* Set the operand to the function object returned to,
|
||||
* to assist optimization passes */
|
||||
ADD_TO_TRACE(uop, oparg, (uintptr_t)func, target);
|
||||
/* Set the operand to the function or code object returned to,
|
||||
* to assist optimization passes. (See _PUSH_FRAME below.)
|
||||
*/
|
||||
if (func != NULL) {
|
||||
operand = (uintptr_t)func;
|
||||
}
|
||||
else if (code != NULL) {
|
||||
operand = (uintptr_t)code | 1;
|
||||
}
|
||||
else {
|
||||
operand = 0;
|
||||
}
|
||||
ADD_TO_TRACE(uop, oparg, operand, target);
|
||||
DPRINTF(2,
|
||||
"Returning to %s (%s:%d) at byte offset %d\n",
|
||||
PyUnicode_AsUTF8(code->co_qualname),
|
||||
|
|
@ -738,10 +751,12 @@ top: // Jump here after _PUSH_FRAME or likely branches
|
|||
// Add one to account for the actual opcode/oparg pair:
|
||||
+ 1;
|
||||
uint32_t func_version = read_u32(&instr[func_version_offset].cache);
|
||||
PyFunctionObject *new_func = _PyFunction_LookupByVersion(func_version);
|
||||
DPRINTF(2, "Function: version=%#x; object=%p\n", (int)func_version, new_func);
|
||||
if (new_func != NULL) {
|
||||
PyCodeObject *new_code = (PyCodeObject *)PyFunction_GET_CODE(new_func);
|
||||
PyCodeObject *new_code = NULL;
|
||||
PyFunctionObject *new_func =
|
||||
_PyFunction_LookupByVersion(func_version, (PyObject **) &new_code);
|
||||
DPRINTF(2, "Function: version=%#x; new_func=%p, new_code=%p\n",
|
||||
(int)func_version, new_func, new_code);
|
||||
if (new_code != NULL) {
|
||||
if (new_code == code) {
|
||||
// Recursive call, bail (we could be here forever).
|
||||
DPRINTF(2, "Bailing on recursive call to %s (%s:%d)\n",
|
||||
|
|
@ -766,9 +781,22 @@ top: // Jump here after _PUSH_FRAME or likely branches
|
|||
instr += _PyOpcode_Caches[_PyOpcode_Deopt[opcode]] + 1;
|
||||
TRACE_STACK_PUSH();
|
||||
_Py_BloomFilter_Add(dependencies, new_code);
|
||||
/* Set the operand to the callee's function object,
|
||||
* to assist optimization passes */
|
||||
ADD_TO_TRACE(uop, oparg, (uintptr_t)new_func, target);
|
||||
/* Set the operand to the callee's function or code object,
|
||||
* to assist optimization passes.
|
||||
* We prefer setting it to the function (for remove_globals())
|
||||
* but if that's not available but the code is available,
|
||||
* use the code, setting the low bit so the optimizer knows.
|
||||
*/
|
||||
if (new_func != NULL) {
|
||||
operand = (uintptr_t)new_func;
|
||||
}
|
||||
else if (new_code != NULL) {
|
||||
operand = (uintptr_t)new_code | 1;
|
||||
}
|
||||
else {
|
||||
operand = 0;
|
||||
}
|
||||
ADD_TO_TRACE(uop, oparg, operand, target);
|
||||
code = new_code;
|
||||
func = new_func;
|
||||
instr = _PyCode_CODE(code);
|
||||
|
|
@ -780,8 +808,8 @@ top: // Jump here after _PUSH_FRAME or likely branches
|
|||
2 * INSTR_IP(instr, code));
|
||||
goto top;
|
||||
}
|
||||
DPRINTF(2, "Bail, new_func == NULL\n");
|
||||
ADD_TO_TRACE(uop, oparg, operand, target);
|
||||
DPRINTF(2, "Bail, new_code == NULL\n");
|
||||
ADD_TO_TRACE(uop, oparg, 0, target);
|
||||
ADD_TO_TRACE(_EXIT_TRACE, 0, 0, 0);
|
||||
goto done;
|
||||
}
|
||||
|
|
@ -1116,7 +1144,7 @@ counter_optimize(
|
|||
int Py_UNUSED(curr_stackentries)
|
||||
)
|
||||
{
|
||||
PyCodeObject *code = (PyCodeObject *)frame->f_executable;
|
||||
PyCodeObject *code = _PyFrame_GetCode(frame);
|
||||
int oparg = instr->op.arg;
|
||||
while (instr->op.code == EXTENDED_ARG) {
|
||||
instr++;
|
||||
|
|
|
|||
|
|
@ -228,7 +228,12 @@ remove_globals(_PyInterpreterFrame *frame, _PyUOpInstruction *buffer,
|
|||
builtins_watched <<= 1;
|
||||
globals_watched <<= 1;
|
||||
function_checked <<= 1;
|
||||
PyFunctionObject *func = (PyFunctionObject *)buffer[pc].operand;
|
||||
uint64_t operand = buffer[pc].operand;
|
||||
if (operand == 0 || (operand & 1)) {
|
||||
// It's either a code object or NULL, so bail
|
||||
return 1;
|
||||
}
|
||||
PyFunctionObject *func = (PyFunctionObject *)operand;
|
||||
if (func == NULL) {
|
||||
return 1;
|
||||
}
|
||||
|
|
@ -251,7 +256,15 @@ remove_globals(_PyInterpreterFrame *frame, _PyUOpInstruction *buffer,
|
|||
builtins_watched >>= 1;
|
||||
globals_watched >>= 1;
|
||||
function_checked >>= 1;
|
||||
PyFunctionObject *func = (PyFunctionObject *)buffer[pc].operand;
|
||||
uint64_t operand = buffer[pc].operand;
|
||||
if (operand == 0 || (operand & 1)) {
|
||||
// It's either a code object or NULL, so bail
|
||||
return 1;
|
||||
}
|
||||
PyFunctionObject *func = (PyFunctionObject *)operand;
|
||||
if (func == NULL) {
|
||||
return 1;
|
||||
}
|
||||
assert(PyFunction_Check(func));
|
||||
function_version = func->func_version;
|
||||
globals = func->func_globals;
|
||||
|
|
@ -522,7 +535,7 @@ remove_unneeded_uops(_PyUOpInstruction *buffer, int buffer_size)
|
|||
static void
|
||||
peephole_opt(_PyInterpreterFrame *frame, _PyUOpInstruction *buffer, int buffer_size)
|
||||
{
|
||||
PyCodeObject *co = (PyCodeObject *)frame->f_executable;
|
||||
PyCodeObject *co = _PyFrame_GetCode(frame);
|
||||
for (int pc = 0; pc < buffer_size; pc++) {
|
||||
int opcode = buffer[pc].opcode;
|
||||
switch(opcode) {
|
||||
|
|
@ -545,11 +558,16 @@ peephole_opt(_PyInterpreterFrame *frame, _PyUOpInstruction *buffer, int buffer_s
|
|||
case _PUSH_FRAME:
|
||||
case _POP_FRAME:
|
||||
{
|
||||
PyFunctionObject *func = (PyFunctionObject *)buffer[pc].operand;
|
||||
if (func == NULL) {
|
||||
uint64_t operand = buffer[pc].operand;
|
||||
if (operand & 1) {
|
||||
co = (PyCodeObject *)(operand & ~1);
|
||||
assert(PyCode_Check(co));
|
||||
}
|
||||
else if (operand == 0) {
|
||||
co = NULL;
|
||||
}
|
||||
else {
|
||||
PyFunctionObject *func = (PyFunctionObject *)operand;
|
||||
assert(PyFunction_Check(func));
|
||||
co = (PyCodeObject *)func->func_code;
|
||||
}
|
||||
|
|
@ -587,7 +605,7 @@ _Py_uop_analyze_and_optimize(
|
|||
peephole_opt(frame, buffer, buffer_size);
|
||||
|
||||
err = optimize_uops(
|
||||
(PyCodeObject *)frame->f_executable, buffer,
|
||||
_PyFrame_GetCode(frame), buffer,
|
||||
buffer_size, curr_stacklen, dependencies);
|
||||
|
||||
if (err == 0) {
|
||||
|
|
|
|||
|
|
@ -543,14 +543,25 @@ dummy_func(void) {
|
|||
|
||||
(void)callable;
|
||||
|
||||
PyFunctionObject *func = (PyFunctionObject *)(this_instr + 2)->operand;
|
||||
DPRINTF(3, "func: %p ", func);
|
||||
if (func == NULL) {
|
||||
DPRINTF(3, "\n");
|
||||
DPRINTF(1, "Missing function\n");
|
||||
goto done;
|
||||
PyCodeObject *co = NULL;
|
||||
assert((this_instr + 2)->opcode == _PUSH_FRAME);
|
||||
uintptr_t push_operand = (this_instr + 2)->operand;
|
||||
if (push_operand & 1) {
|
||||
co = (PyCodeObject *)(push_operand & ~1);
|
||||
DPRINTF(3, "code=%p ", co);
|
||||
assert(PyCode_Check(co));
|
||||
}
|
||||
else {
|
||||
PyFunctionObject *func = (PyFunctionObject *)push_operand;
|
||||
DPRINTF(3, "func=%p ", func);
|
||||
if (func == NULL) {
|
||||
DPRINTF(3, "\n");
|
||||
DPRINTF(1, "Missing function\n");
|
||||
goto done;
|
||||
}
|
||||
co = (PyCodeObject *)func->func_code;
|
||||
DPRINTF(3, "code=%p ", co);
|
||||
}
|
||||
PyCodeObject *co = (PyCodeObject *)func->func_code;
|
||||
|
||||
assert(self_or_null != NULL);
|
||||
assert(args != NULL);
|
||||
|
|
|
|||
25
Python/optimizer_cases.c.h
generated
25
Python/optimizer_cases.c.h
generated
|
|
@ -1596,14 +1596,25 @@
|
|||
callable = stack_pointer[-2 - oparg];
|
||||
int argcount = oparg;
|
||||
(void)callable;
|
||||
PyFunctionObject *func = (PyFunctionObject *)(this_instr + 2)->operand;
|
||||
DPRINTF(3, "func: %p ", func);
|
||||
if (func == NULL) {
|
||||
DPRINTF(3, "\n");
|
||||
DPRINTF(1, "Missing function\n");
|
||||
goto done;
|
||||
PyCodeObject *co = NULL;
|
||||
assert((this_instr + 2)->opcode == _PUSH_FRAME);
|
||||
uintptr_t push_operand = (this_instr + 2)->operand;
|
||||
if (push_operand & 1) {
|
||||
co = (PyCodeObject *)(push_operand & ~1);
|
||||
DPRINTF(3, "code=%p ", co);
|
||||
assert(PyCode_Check(co));
|
||||
}
|
||||
else {
|
||||
PyFunctionObject *func = (PyFunctionObject *)push_operand;
|
||||
DPRINTF(3, "func=%p ", func);
|
||||
if (func == NULL) {
|
||||
DPRINTF(3, "\n");
|
||||
DPRINTF(1, "Missing function\n");
|
||||
goto done;
|
||||
}
|
||||
co = (PyCodeObject *)func->func_code;
|
||||
DPRINTF(3, "code=%p ", co);
|
||||
}
|
||||
PyCodeObject *co = (PyCodeObject *)func->func_code;
|
||||
assert(self_or_null != NULL);
|
||||
assert(args != NULL);
|
||||
if (sym_is_not_null(self_or_null)) {
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue