#include "Python.h" #include "opcode.h" #include "pycore_interp.h" #include "pycore_bitutils.h" // _Py_popcount32() #include "pycore_object.h" // _PyObject_GC_UNTRACK() #include "pycore_opcode_metadata.h" // _PyOpcode_OpName[] #include "pycore_opcode_utils.h" // MAX_REAL_OPCODE #include "pycore_optimizer.h" // _Py_uop_analyze_and_optimize() #include "pycore_pystate.h" // _PyInterpreterState_GET() #include "pycore_uop_ids.h" #include "pycore_jit.h" #include "cpython/optimizer.h" #include #include #include #define NEED_OPCODE_METADATA #include "pycore_uop_metadata.h" // Uop tables #undef NEED_OPCODE_METADATA #define MAX_EXECUTORS_SIZE 256 static bool has_space_for_executor(PyCodeObject *code, _Py_CODEUNIT *instr) { if (instr->op.code == ENTER_EXECUTOR) { return true; } if (code->co_executors == NULL) { return true; } return code->co_executors->size < MAX_EXECUTORS_SIZE; } static int32_t get_index_for_executor(PyCodeObject *code, _Py_CODEUNIT *instr) { if (instr->op.code == ENTER_EXECUTOR) { return instr->op.arg; } _PyExecutorArray *old = code->co_executors; int size = 0; int capacity = 0; if (old != NULL) { size = old->size; capacity = old->capacity; assert(size < MAX_EXECUTORS_SIZE); } assert(size <= capacity); if (size == capacity) { /* Array is full. Grow array */ int new_capacity = capacity ? capacity * 2 : 4; _PyExecutorArray *new = PyMem_Realloc( old, offsetof(_PyExecutorArray, executors) + new_capacity * sizeof(_PyExecutorObject *)); if (new == NULL) { return -1; } new->capacity = new_capacity; new->size = size; code->co_executors = new; } assert(size < code->co_executors->capacity); return size; } static void insert_executor(PyCodeObject *code, _Py_CODEUNIT *instr, int index, _PyExecutorObject *executor) { Py_INCREF(executor); if (instr->op.code == ENTER_EXECUTOR) { assert(index == instr->op.arg); _Py_ExecutorClear(code->co_executors->executors[index]); } else { assert(code->co_executors->size == index); assert(code->co_executors->capacity > index); code->co_executors->size++; } executor->vm_data.opcode = instr->op.code; executor->vm_data.oparg = instr->op.arg; executor->vm_data.code = code; executor->vm_data.index = (int)(instr - _PyCode_CODE(code)); code->co_executors->executors[index] = executor; assert(index < MAX_EXECUTORS_SIZE); instr->op.code = ENTER_EXECUTOR; instr->op.arg = index; } int PyUnstable_Replace_Executor(PyCodeObject *code, _Py_CODEUNIT *instr, _PyExecutorObject *new) { if (instr->op.code != ENTER_EXECUTOR) { PyErr_Format(PyExc_ValueError, "No executor to replace"); return -1; } int index = instr->op.arg; assert(index >= 0); insert_executor(code, instr, index, new); return 0; } static int never_optimize( _PyOptimizerObject* self, _PyInterpreterFrame *frame, _Py_CODEUNIT *instr, _PyExecutorObject **exec, int Py_UNUSED(stack_entries)) { /* Although it should be benign for this to be called, * it shouldn't happen, so fail in debug builds. */ assert(0 && "never optimize should never be called"); return 0; } PyTypeObject _PyDefaultOptimizer_Type = { PyVarObject_HEAD_INIT(&PyType_Type, 0) .tp_name = "noop_optimizer", .tp_basicsize = sizeof(_PyOptimizerObject), .tp_itemsize = 0, .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_DISALLOW_INSTANTIATION, }; static _PyOptimizerObject _PyOptimizer_Default = { PyObject_HEAD_INIT(&_PyDefaultOptimizer_Type) .optimize = never_optimize, .resume_threshold = OPTIMIZER_UNREACHABLE_THRESHOLD, .backedge_threshold = OPTIMIZER_UNREACHABLE_THRESHOLD, .side_threshold = OPTIMIZER_UNREACHABLE_THRESHOLD, }; static uint32_t shift_and_offset_threshold(uint32_t threshold) { return (threshold << OPTIMIZER_BITS_IN_COUNTER) + (1 << 15); } _PyOptimizerObject * PyUnstable_GetOptimizer(void) { PyInterpreterState *interp = _PyInterpreterState_GET(); assert(interp->optimizer_backedge_threshold == shift_and_offset_threshold(interp->optimizer->backedge_threshold)); assert(interp->optimizer_resume_threshold == shift_and_offset_threshold(interp->optimizer->resume_threshold)); if (interp->optimizer == &_PyOptimizer_Default) { return NULL; } Py_INCREF(interp->optimizer); return interp->optimizer; } static _PyExecutorObject * make_executor_from_uops(_PyUOpInstruction *buffer, const _PyBloomFilter *dependencies); static int init_cold_exit_executor(_PyExecutorObject *executor, int oparg); static int cold_exits_initialized = 0; static _PyExecutorObject COLD_EXITS[UOP_MAX_TRACE_LENGTH] = { 0 }; static const _PyBloomFilter EMPTY_FILTER = { 0 }; _PyOptimizerObject * _Py_SetOptimizer(PyInterpreterState *interp, _PyOptimizerObject *optimizer) { if (optimizer == NULL) { optimizer = &_PyOptimizer_Default; } else if (cold_exits_initialized == 0) { cold_exits_initialized = 1; for (int i = 0; i < UOP_MAX_TRACE_LENGTH; i++) { if (init_cold_exit_executor(&COLD_EXITS[i], i)) { return NULL; } } } _PyOptimizerObject *old = interp->optimizer; if (old == NULL) { old = &_PyOptimizer_Default; } Py_INCREF(optimizer); interp->optimizer = optimizer; interp->optimizer_backedge_threshold = shift_and_offset_threshold(optimizer->backedge_threshold); interp->optimizer_resume_threshold = shift_and_offset_threshold(optimizer->resume_threshold); interp->optimizer_side_threshold = optimizer->side_threshold; if (optimizer == &_PyOptimizer_Default) { assert(interp->optimizer_backedge_threshold > (1 << 16)); assert(interp->optimizer_resume_threshold > (1 << 16)); } return old; } int PyUnstable_SetOptimizer(_PyOptimizerObject *optimizer) { PyInterpreterState *interp = _PyInterpreterState_GET(); _PyOptimizerObject *old = _Py_SetOptimizer(interp, optimizer); Py_XDECREF(old); return old == NULL ? -1 : 0; } /* Returns 1 if optimized, 0 if not optimized, and -1 for an error. * If optimized, *executor_ptr contains a new reference to the executor */ int _PyOptimizer_Optimize( _PyInterpreterFrame *frame, _Py_CODEUNIT *start, PyObject **stack_pointer, _PyExecutorObject **executor_ptr) { PyCodeObject *code = (PyCodeObject *)frame->f_executable; assert(PyCode_Check(code)); PyInterpreterState *interp = _PyInterpreterState_GET(); if (!has_space_for_executor(code, start)) { return 0; } _PyOptimizerObject *opt = interp->optimizer; int err = opt->optimize(opt, frame, start, executor_ptr, (int)(stack_pointer - _PyFrame_Stackbase(frame))); if (err <= 0) { return err; } assert(*executor_ptr != NULL); int index = get_index_for_executor(code, start); if (index < 0) { /* Out of memory. Don't raise and assume that the * error will show up elsewhere. * * If an optimizer has already produced an executor, * it might get confused by the executor disappearing, * but there is not much we can do about that here. */ Py_DECREF(*executor_ptr); return 0; } insert_executor(code, start, index, *executor_ptr); assert((*executor_ptr)->vm_data.valid); return 1; } _PyExecutorObject * PyUnstable_GetExecutor(PyCodeObject *code, int offset) { int code_len = (int)Py_SIZE(code); for (int i = 0 ; i < code_len;) { if (_PyCode_CODE(code)[i].op.code == ENTER_EXECUTOR && i*2 == offset) { int oparg = _PyCode_CODE(code)[i].op.arg; _PyExecutorObject *res = code->co_executors->executors[oparg]; Py_INCREF(res); return res; } i += _PyInstruction_GetLength(code, i); } PyErr_SetString(PyExc_ValueError, "no executor at given byte offset"); return NULL; } static PyObject * is_valid(PyObject *self, PyObject *Py_UNUSED(ignored)) { return PyBool_FromLong(((_PyExecutorObject *)self)->vm_data.valid); } static PyObject * get_opcode(PyObject *self, PyObject *Py_UNUSED(ignored)) { return PyLong_FromUnsignedLong(((_PyExecutorObject *)self)->vm_data.opcode); } static PyObject * get_oparg(PyObject *self, PyObject *Py_UNUSED(ignored)) { return PyLong_FromUnsignedLong(((_PyExecutorObject *)self)->vm_data.oparg); } static PyMethodDef executor_methods[] = { { "is_valid", is_valid, METH_NOARGS, NULL }, { "get_opcode", get_opcode, METH_NOARGS, NULL }, { "get_oparg", get_oparg, METH_NOARGS, NULL }, { NULL, NULL }, }; ///////////////////// Experimental UOp Optimizer ///////////////////// static void uop_dealloc(_PyExecutorObject *self) { _PyObject_GC_UNTRACK(self); _Py_ExecutorClear(self); #ifdef _Py_JIT _PyJIT_Free(self); #endif PyObject_GC_Del(self); } const char * _PyUOpName(int index) { if (index < 0 || index > MAX_UOP_ID) { return NULL; } return _PyOpcode_uop_name[index]; } #ifdef Py_DEBUG void _PyUOpPrint(const _PyUOpInstruction *uop) { const char *name = _PyUOpName(uop->opcode); if (name == NULL) { printf("", uop->opcode); } else { printf("%s", name); } printf(" (%d, target=%d, operand=%" PRIx64 ")", uop->oparg, uop->target, (uint64_t)uop->operand); } #endif static Py_ssize_t uop_len(_PyExecutorObject *self) { return self->code_size; } static PyObject * uop_item(_PyExecutorObject *self, Py_ssize_t index) { Py_ssize_t len = uop_len(self); if (index < 0 || index >= len) { PyErr_SetNone(PyExc_IndexError); return NULL; } const char *name = _PyUOpName(self->trace[index].opcode); if (name == NULL) { name = ""; } PyObject *oname = _PyUnicode_FromASCII(name, strlen(name)); if (oname == NULL) { return NULL; } PyObject *oparg = PyLong_FromUnsignedLong(self->trace[index].oparg); if (oparg == NULL) { Py_DECREF(oname); return NULL; } PyObject *target = PyLong_FromUnsignedLong(self->trace[index].target); if (oparg == NULL) { Py_DECREF(oparg); Py_DECREF(oname); return NULL; } PyObject *operand = PyLong_FromUnsignedLongLong(self->trace[index].operand); if (operand == NULL) { Py_DECREF(target); Py_DECREF(oparg); Py_DECREF(oname); return NULL; } PyObject *args[4] = { oname, oparg, target, operand }; return _PyTuple_FromArraySteal(args, 4); } PySequenceMethods uop_as_sequence = { .sq_length = (lenfunc)uop_len, .sq_item = (ssizeargfunc)uop_item, }; static int executor_clear(PyObject *o) { _Py_ExecutorClear((_PyExecutorObject *)o); return 0; } static int executor_traverse(PyObject *o, visitproc visit, void *arg) { _PyExecutorObject *executor = (_PyExecutorObject *)o; for (uint32_t i = 0; i < executor->exit_count; i++) { Py_VISIT(executor->exits[i].executor); } return 0; } PyTypeObject _PyUOpExecutor_Type = { PyVarObject_HEAD_INIT(&PyType_Type, 0) .tp_name = "uop_executor", .tp_basicsize = offsetof(_PyExecutorObject, exits), .tp_itemsize = 1, .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_DISALLOW_INSTANTIATION | Py_TPFLAGS_HAVE_GC, .tp_dealloc = (destructor)uop_dealloc, .tp_as_sequence = &uop_as_sequence, .tp_methods = executor_methods, .tp_traverse = executor_traverse, .tp_clear = executor_clear, }; /* TO DO -- Generate these tables */ static const uint16_t _PyUOp_Replacements[MAX_UOP_ID + 1] = { [_ITER_JUMP_RANGE] = _GUARD_NOT_EXHAUSTED_RANGE, [_ITER_JUMP_LIST] = _GUARD_NOT_EXHAUSTED_LIST, [_ITER_JUMP_TUPLE] = _GUARD_NOT_EXHAUSTED_TUPLE, [_FOR_ITER] = _FOR_ITER_TIER_TWO, }; static const uint16_t BRANCH_TO_GUARD[4][2] = { [POP_JUMP_IF_FALSE - POP_JUMP_IF_FALSE][0] = _GUARD_IS_TRUE_POP, [POP_JUMP_IF_FALSE - POP_JUMP_IF_FALSE][1] = _GUARD_IS_FALSE_POP, [POP_JUMP_IF_TRUE - POP_JUMP_IF_FALSE][0] = _GUARD_IS_FALSE_POP, [POP_JUMP_IF_TRUE - POP_JUMP_IF_FALSE][1] = _GUARD_IS_TRUE_POP, [POP_JUMP_IF_NONE - POP_JUMP_IF_FALSE][0] = _GUARD_IS_NOT_NONE_POP, [POP_JUMP_IF_NONE - POP_JUMP_IF_FALSE][1] = _GUARD_IS_NONE_POP, [POP_JUMP_IF_NOT_NONE - POP_JUMP_IF_FALSE][0] = _GUARD_IS_NONE_POP, [POP_JUMP_IF_NOT_NONE - POP_JUMP_IF_FALSE][1] = _GUARD_IS_NOT_NONE_POP, }; #define CONFIDENCE_RANGE 1000 #define CONFIDENCE_CUTOFF 333 #ifdef Py_DEBUG #define DPRINTF(level, ...) \ if (lltrace >= (level)) { printf(__VA_ARGS__); } #else #define DPRINTF(level, ...) #endif // Beware: Macro arg order differs from struct member order #ifdef Py_DEBUG #define ADD_TO_TRACE(OPCODE, OPARG, OPERAND, TARGET) \ assert(trace_length < max_length); \ trace[trace_length].opcode = (OPCODE); \ trace[trace_length].oparg = (OPARG); \ trace[trace_length].target = (TARGET); \ trace[trace_length].operand = (OPERAND); \ if (lltrace >= 2) { \ printf("%4d ADD_TO_TRACE: ", trace_length); \ _PyUOpPrint(&trace[trace_length]); \ printf("\n"); \ } \ trace_length++; #else #define ADD_TO_TRACE(OPCODE, OPARG, OPERAND, TARGET) \ assert(trace_length < max_length); \ trace[trace_length].opcode = (OPCODE); \ trace[trace_length].oparg = (OPARG); \ trace[trace_length].target = (TARGET); \ trace[trace_length].operand = (OPERAND); \ trace_length++; #endif #define INSTR_IP(INSTR, CODE) \ ((uint32_t)((INSTR) - ((_Py_CODEUNIT *)(CODE)->co_code_adaptive))) // Reserve space for n uops #define RESERVE_RAW(n, opname) \ if (trace_length + (n) > max_length) { \ DPRINTF(2, "No room for %s (need %d, got %d)\n", \ (opname), (n), max_length - trace_length); \ OPT_STAT_INC(trace_too_long); \ goto done; \ } // Reserve space for N uops, plus 3 for _SET_IP, _CHECK_VALIDITY and _EXIT_TRACE #define RESERVE(needed) RESERVE_RAW((needed) + 3, _PyUOpName(opcode)) // Trace stack operations (used by _PUSH_FRAME, _POP_FRAME) #define TRACE_STACK_PUSH() \ if (trace_stack_depth >= TRACE_STACK_SIZE) { \ DPRINTF(2, "Trace stack overflow\n"); \ OPT_STAT_INC(trace_stack_overflow); \ ADD_TO_TRACE(_EXIT_TRACE, 0, 0, 0); \ goto done; \ } \ assert(func->func_code == (PyObject *)code); \ trace_stack[trace_stack_depth].func = func; \ trace_stack[trace_stack_depth].instr = instr; \ trace_stack_depth++; #define TRACE_STACK_POP() \ if (trace_stack_depth <= 0) { \ Py_FatalError("Trace stack underflow\n"); \ } \ trace_stack_depth--; \ func = trace_stack[trace_stack_depth].func; \ code = (PyCodeObject *)trace_stack[trace_stack_depth].func->func_code; \ instr = trace_stack[trace_stack_depth].instr; /* Returns 1 on success, * 0 if it failed to produce a worthwhile trace, * and -1 on an error. */ static int translate_bytecode_to_trace( _PyInterpreterFrame *frame, _Py_CODEUNIT *instr, _PyUOpInstruction *trace, int buffer_size, _PyBloomFilter *dependencies) { bool progress_needed = true; PyCodeObject *code = (PyCodeObject *)frame->f_executable; PyFunctionObject *func = (PyFunctionObject *)frame->f_funcobj; assert(PyFunction_Check(func)); PyCodeObject *initial_code = code; _Py_BloomFilter_Add(dependencies, initial_code); _Py_CODEUNIT *initial_instr = instr; int trace_length = 0; int max_length = buffer_size; struct { PyFunctionObject *func; _Py_CODEUNIT *instr; } trace_stack[TRACE_STACK_SIZE]; int trace_stack_depth = 0; int confidence = CONFIDENCE_RANGE; // Adjusted by branch instructions #ifdef Py_DEBUG char *python_lltrace = Py_GETENV("PYTHON_LLTRACE"); int lltrace = 0; if (python_lltrace != NULL && *python_lltrace >= '0') { lltrace = *python_lltrace - '0'; // TODO: Parse an int and all that } #endif DPRINTF(4, "Optimizing %s (%s:%d) at byte offset %d\n", PyUnicode_AsUTF8(code->co_qualname), PyUnicode_AsUTF8(code->co_filename), code->co_firstlineno, 2 * INSTR_IP(initial_instr, code)); uint32_t target = 0; top: // Jump here after _PUSH_FRAME or likely branches for (;;) { target = INSTR_IP(instr, code); RESERVE_RAW(2, "epilogue"); // Always need space for _SET_IP, _CHECK_VALIDITY and _EXIT_TRACE ADD_TO_TRACE(_CHECK_VALIDITY_AND_SET_IP, 0, (uintptr_t)instr, target); uint32_t opcode = instr->op.code; uint32_t oparg = instr->op.arg; uint32_t extended = 0; DPRINTF(3, "%d: %s(%d)\n", target, _PyOpcode_OpName[opcode], oparg); if (opcode == ENTER_EXECUTOR) { assert(oparg < 256); _PyExecutorObject *executor = code->co_executors->executors[oparg]; opcode = executor->vm_data.opcode; DPRINTF(2, " * ENTER_EXECUTOR -> %s\n", _PyOpcode_OpName[opcode]); oparg = executor->vm_data.oparg; } if (opcode == EXTENDED_ARG) { instr++; extended = 1; opcode = instr->op.code; oparg = (oparg << 8) | instr->op.arg; if (opcode == EXTENDED_ARG) { instr--; goto done; } } assert(opcode != ENTER_EXECUTOR && opcode != EXTENDED_ARG); /* Special case the first instruction, * so that we can guarantee forward progress */ if (progress_needed) { progress_needed = false; if (opcode == JUMP_BACKWARD || opcode == JUMP_BACKWARD_NO_INTERRUPT) { instr += 1 + _PyOpcode_Caches[opcode] - (int32_t)oparg; initial_instr = instr; continue; } else { if (OPCODE_HAS_DEOPT(opcode)) { opcode = _PyOpcode_Deopt[opcode]; } assert(!OPCODE_HAS_DEOPT(opcode)); } } switch (opcode) { case POP_JUMP_IF_NONE: case POP_JUMP_IF_NOT_NONE: case POP_JUMP_IF_FALSE: case POP_JUMP_IF_TRUE: { RESERVE(1); int counter = instr[1].cache; int bitcount = _Py_popcount32(counter); int jump_likely = bitcount > 8; /* If bitcount is 8 (half the jumps were taken), adjust confidence by 50%. If it's 16 or 0 (all or none were taken), adjust by 10% (since the future is still somewhat uncertain). For values in between, adjust proportionally. */ if (jump_likely) { confidence = confidence * (bitcount + 2) / 20; } else { confidence = confidence * (18 - bitcount) / 20; } uint32_t uopcode = BRANCH_TO_GUARD[opcode - POP_JUMP_IF_FALSE][jump_likely]; DPRINTF(2, "%d: %s(%d): counter=%x, bitcount=%d, likely=%d, confidence=%d, uopcode=%s\n", target, _PyOpcode_OpName[opcode], oparg, counter, bitcount, jump_likely, confidence, _PyUOpName(uopcode)); if (confidence < CONFIDENCE_CUTOFF) { DPRINTF(2, "Confidence too low (%d < %d)\n", confidence, CONFIDENCE_CUTOFF); OPT_STAT_INC(low_confidence); goto done; } _Py_CODEUNIT *next_instr = instr + 1 + _PyOpcode_Caches[_PyOpcode_Deopt[opcode]]; _Py_CODEUNIT *target_instr = next_instr + oparg; if (jump_likely) { DPRINTF(2, "Jump likely (%x = %d bits), continue at byte offset %d\n", instr[1].cache, bitcount, 2 * INSTR_IP(target_instr, code)); instr = target_instr; ADD_TO_TRACE(uopcode, max_length, 0, INSTR_IP(next_instr, code)); goto top; } ADD_TO_TRACE(uopcode, max_length, 0, INSTR_IP(target_instr, code)); break; } case JUMP_BACKWARD: case JUMP_BACKWARD_NO_INTERRUPT: { _Py_CODEUNIT *target = instr + 1 + _PyOpcode_Caches[opcode] - (int)oparg; if (target == initial_instr) { /* We have looped round to the start */ RESERVE(1); ADD_TO_TRACE(_JUMP_TO_TOP, 0, 0, 0); } else { OPT_STAT_INC(inner_loop); DPRINTF(2, "JUMP_BACKWARD not to top ends trace\n"); } goto done; } case JUMP_FORWARD: { RESERVE(0); // This will emit two _SET_IP instructions; leave it to the optimizer instr += oparg; break; } default: { const struct opcode_macro_expansion *expansion = &_PyOpcode_macro_expansion[opcode]; if (expansion->nuops > 0) { // Reserve space for nuops (+ _SET_IP + _EXIT_TRACE) int nuops = expansion->nuops; RESERVE(nuops); if (expansion->uops[nuops-1].uop == _POP_FRAME) { // Check for trace stack underflow now: // We can't bail e.g. in the middle of // LOAD_CONST + _POP_FRAME. if (trace_stack_depth == 0) { DPRINTF(2, "Trace stack underflow\n"); OPT_STAT_INC(trace_stack_underflow); goto done; } } uint32_t orig_oparg = oparg; // For OPARG_TOP/BOTTOM for (int i = 0; i < nuops; i++) { oparg = orig_oparg; uint32_t uop = expansion->uops[i].uop; uint64_t operand = 0; // Add one to account for the actual opcode/oparg pair: int offset = expansion->uops[i].offset + 1; switch (expansion->uops[i].size) { case OPARG_FULL: assert(opcode != JUMP_BACKWARD_NO_INTERRUPT && opcode != JUMP_BACKWARD); break; case OPARG_CACHE_1: operand = read_u16(&instr[offset].cache); break; case OPARG_CACHE_2: operand = read_u32(&instr[offset].cache); break; case OPARG_CACHE_4: operand = read_u64(&instr[offset].cache); break; case OPARG_TOP: // First half of super-instr oparg = orig_oparg >> 4; break; case OPARG_BOTTOM: // Second half of super-instr oparg = orig_oparg & 0xF; break; case OPARG_SAVE_RETURN_OFFSET: // op=_SAVE_RETURN_OFFSET; oparg=return_offset oparg = offset; assert(uop == _SAVE_RETURN_OFFSET); break; case OPARG_REPLACED: uop = _PyUOp_Replacements[uop]; assert(uop != 0); if (uop == _FOR_ITER_TIER_TWO) { target += 1 + INLINE_CACHE_ENTRIES_FOR_ITER + oparg + 2 + extended; assert(_PyCode_CODE(code)[target-2].op.code == END_FOR || _PyCode_CODE(code)[target-2].op.code == INSTRUMENTED_END_FOR); assert(_PyCode_CODE(code)[target-1].op.code == POP_TOP); } break; default: fprintf(stderr, "opcode=%d, oparg=%d; nuops=%d, i=%d; size=%d, offset=%d\n", opcode, oparg, nuops, i, expansion->uops[i].size, expansion->uops[i].offset); Py_FatalError("garbled expansion"); } ADD_TO_TRACE(uop, oparg, operand, target); if (uop == _POP_FRAME) { TRACE_STACK_POP(); /* Set the operand to the function object returned to, * to assist optimization passes */ trace[trace_length-1].operand = (uintptr_t)func; DPRINTF(2, "Returning to %s (%s:%d) at byte offset %d\n", PyUnicode_AsUTF8(code->co_qualname), PyUnicode_AsUTF8(code->co_filename), code->co_firstlineno, 2 * INSTR_IP(instr, code)); goto top; } if (uop == _PUSH_FRAME) { assert(i + 1 == nuops); int func_version_offset = offsetof(_PyCallCache, func_version)/sizeof(_Py_CODEUNIT) // 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(3, "Function object: %p\n", func); if (new_func != NULL) { PyCodeObject *new_code = (PyCodeObject *)PyFunction_GET_CODE(new_func); if (new_code == code) { // Recursive call, bail (we could be here forever). DPRINTF(2, "Bailing on recursive call to %s (%s:%d)\n", PyUnicode_AsUTF8(new_code->co_qualname), PyUnicode_AsUTF8(new_code->co_filename), new_code->co_firstlineno); OPT_STAT_INC(recursive_call); ADD_TO_TRACE(_EXIT_TRACE, 0, 0, 0); goto done; } if (new_code->co_version != func_version) { // func.__code__ was updated. // Perhaps it may happen again, so don't bother tracing. // TODO: Reason about this -- is it better to bail or not? DPRINTF(2, "Bailing because co_version != func_version\n"); ADD_TO_TRACE(_EXIT_TRACE, 0, 0, 0); goto done; } // Increment IP to the return address instr += _PyOpcode_Caches[_PyOpcode_Deopt[opcode]] + 1; TRACE_STACK_PUSH(); _Py_BloomFilter_Add(dependencies, new_code); /* Set the operand to the callee's code object, * to assist optimization passes */ trace[trace_length-1].operand = (uintptr_t)new_func; code = new_code; func = new_func; instr = _PyCode_CODE(code); DPRINTF(2, "Continuing in %s (%s:%d) at byte offset %d\n", PyUnicode_AsUTF8(code->co_qualname), PyUnicode_AsUTF8(code->co_filename), code->co_firstlineno, 2 * INSTR_IP(instr, code)); goto top; } ADD_TO_TRACE(_EXIT_TRACE, 0, 0, 0); goto done; } } break; } DPRINTF(2, "Unsupported opcode %s\n", _PyOpcode_OpName[opcode]); OPT_UNSUPPORTED_OPCODE(opcode); goto done; // Break out of loop } // End default } // End switch (opcode) instr++; // Add cache size for opcode instr += _PyOpcode_Caches[_PyOpcode_Deopt[opcode]]; } // End for (;;) done: while (trace_stack_depth > 0) { TRACE_STACK_POP(); } assert(code == initial_code); // Skip short traces like _SET_IP, LOAD_FAST, _SET_IP, _EXIT_TRACE if (progress_needed || trace_length < 5) { OPT_STAT_INC(trace_too_short); DPRINTF(4, "No trace for %s (%s:%d) at byte offset %d\n", PyUnicode_AsUTF8(code->co_qualname), PyUnicode_AsUTF8(code->co_filename), code->co_firstlineno, 2 * INSTR_IP(initial_instr, code)); return 0; } ADD_TO_TRACE(_EXIT_TRACE, 0, 0, target); DPRINTF(1, "Created a trace for %s (%s:%d) at byte offset %d -- length %d\n", PyUnicode_AsUTF8(code->co_qualname), PyUnicode_AsUTF8(code->co_filename), code->co_firstlineno, 2 * INSTR_IP(initial_instr, code), trace_length); OPT_HIST(trace_length + buffer_size - max_length, trace_length_hist); return 1; } #undef RESERVE #undef RESERVE_RAW #undef INSTR_IP #undef ADD_TO_TRACE #undef DPRINTF #define UNSET_BIT(array, bit) (array[(bit)>>5] &= ~(1<<((bit)&31))) #define SET_BIT(array, bit) (array[(bit)>>5] |= (1<<((bit)&31))) #define BIT_IS_SET(array, bit) (array[(bit)>>5] & (1<<((bit)&31))) /* Count the number of used uops, and mark them in the bit vector `used`. * This can be done in a single pass using simple reachability analysis, * as there are no backward jumps. * NOPs are excluded from the count. */ static int compute_used(_PyUOpInstruction *buffer, uint32_t *used, int *exit_count_ptr) { int count = 0; int exit_count = 0; SET_BIT(used, 0); for (int i = 0; i < UOP_MAX_TRACE_LENGTH; i++) { if (!BIT_IS_SET(used, i)) { continue; } count++; int opcode = buffer[i].opcode; if (_PyUop_Flags[opcode] & HAS_EXIT_FLAG) { exit_count++; } if (opcode == _JUMP_TO_TOP || opcode == _EXIT_TRACE) { continue; } /* All other micro-ops fall through, so i+1 is reachable */ SET_BIT(used, i+1); assert(opcode <= MAX_UOP_ID); if (_PyUop_Flags[opcode] & HAS_JUMP_FLAG) { /* Mark target as reachable */ SET_BIT(used, buffer[i].oparg); } if (opcode == NOP) { count--; UNSET_BIT(used, i); } } *exit_count_ptr = exit_count; return count; } /* Executor side exits */ static _PyExecutorObject * allocate_executor(int exit_count, int length) { int size = exit_count*sizeof(_PyExitData) + length*sizeof(_PyUOpInstruction); _PyExecutorObject *res = PyObject_GC_NewVar(_PyExecutorObject, &_PyUOpExecutor_Type, size); if (res == NULL) { return NULL; } res->trace = (_PyUOpInstruction *)(res->exits + exit_count); res->code_size = length; res->exit_count = exit_count; return res; } /* Makes an executor from a buffer of uops. * Account for the buffer having gaps and NOPs by computing a "used" * bit vector and only copying the used uops. Here "used" means reachable * and not a NOP. */ static _PyExecutorObject * make_executor_from_uops(_PyUOpInstruction *buffer, const _PyBloomFilter *dependencies) { uint32_t used[(UOP_MAX_TRACE_LENGTH + 31)/32] = { 0 }; int exit_count; int length = compute_used(buffer, used, &exit_count); length += 1; // For _START_EXECUTOR _PyExecutorObject *executor = allocate_executor(exit_count, length); if (executor == NULL) { return NULL; } /* Initialize exits */ for (int i = 0; i < exit_count; i++) { executor->exits[i].executor = &COLD_EXITS[i]; executor->exits[i].temperature = 0; } int next_exit = exit_count-1; _PyUOpInstruction *dest = (_PyUOpInstruction *)&executor->trace[length-1]; /* Scan backwards, so that we see the destinations of jumps before the jumps themselves. */ for (int i = UOP_MAX_TRACE_LENGTH-1; i >= 0; i--) { if (!BIT_IS_SET(used, i)) { continue; } *dest = buffer[i]; int opcode = buffer[i].opcode; if (opcode == _POP_JUMP_IF_FALSE || opcode == _POP_JUMP_IF_TRUE) { /* The oparg of the target will already have been set to its new offset */ int oparg = dest->oparg; dest->oparg = buffer[oparg].oparg; } if (_PyUop_Flags[opcode] & HAS_EXIT_FLAG) { executor->exits[next_exit].target = buffer[i].target; dest->exit_index = next_exit; next_exit--; } /* Set the oparg to be the destination offset, * so that we can set the oparg of earlier jumps correctly. */ buffer[i].oparg = (uint16_t)(dest - executor->trace); dest--; } assert(next_exit == -1); assert(dest == executor->trace); dest->opcode = _START_EXECUTOR; dest->operand = (uintptr_t)executor; _Py_ExecutorInit(executor, dependencies); #ifdef Py_DEBUG char *python_lltrace = Py_GETENV("PYTHON_LLTRACE"); int lltrace = 0; if (python_lltrace != NULL && *python_lltrace >= '0') { lltrace = *python_lltrace - '0'; // TODO: Parse an int and all that } if (lltrace >= 2) { printf("Optimized executor (length %d):\n", length); for (int i = 0; i < length; i++) { printf("%4d OPTIMIZED: ", i); _PyUOpPrint(&executor->trace[i]); printf("\n"); } } #endif #ifdef _Py_JIT executor->jit_code = NULL; executor->jit_size = 0; if (_PyJIT_Compile(executor, executor->trace, length)) { Py_DECREF(executor); return NULL; } #endif _PyObject_GC_TRACK(executor); return executor; } static int init_cold_exit_executor(_PyExecutorObject *executor, int oparg) { _Py_SetImmortal(executor); Py_SET_TYPE(executor, &_PyUOpExecutor_Type); executor->trace = (_PyUOpInstruction *)executor->exits; executor->code_size = 1; executor->exit_count = 0; _PyUOpInstruction *inst = (_PyUOpInstruction *)&executor->trace[0]; inst->opcode = _COLD_EXIT; inst->oparg = oparg; executor->vm_data.valid = true; for (int i = 0; i < BLOOM_FILTER_WORDS; i++) { assert(executor->vm_data.bloom.bits[i] == 0); } #ifdef _Py_JIT executor->jit_code = NULL; executor->jit_size = 0; if (_PyJIT_Compile(executor, executor->trace, 1)) { return -1; } #endif return 0; } static int uop_optimize( _PyOptimizerObject *self, _PyInterpreterFrame *frame, _Py_CODEUNIT *instr, _PyExecutorObject **exec_ptr, int curr_stackentries) { _PyBloomFilter dependencies; _Py_BloomFilter_Init(&dependencies); _PyUOpInstruction buffer[UOP_MAX_TRACE_LENGTH]; int err = translate_bytecode_to_trace(frame, instr, buffer, UOP_MAX_TRACE_LENGTH, &dependencies); if (err <= 0) { // Error or nothing translated return err; } OPT_STAT_INC(traces_created); char *env_var = Py_GETENV("PYTHON_UOPS_OPTIMIZE"); if (env_var == NULL || *env_var == '\0' || *env_var > '0') { err = _Py_uop_analyze_and_optimize(frame, buffer, UOP_MAX_TRACE_LENGTH, curr_stackentries, &dependencies); if (err <= 0) { return err; } } assert(err == 1); /* Fix up */ for (int pc = 0; pc < UOP_MAX_TRACE_LENGTH; pc++) { int opcode = buffer[pc].opcode; int oparg = buffer[pc].oparg; if (_PyUop_Flags[opcode] & HAS_OPARG_AND_1_FLAG) { buffer[pc].opcode = opcode + 1 + (oparg & 1); } else if (oparg < _PyUop_Replication[opcode]) { buffer[pc].opcode = opcode + oparg + 1; } else if (opcode == _JUMP_TO_TOP || opcode == _EXIT_TRACE) { break; } assert(_PyOpcode_uop_name[buffer[pc].opcode]); assert(strncmp(_PyOpcode_uop_name[buffer[pc].opcode], _PyOpcode_uop_name[opcode], strlen(_PyOpcode_uop_name[opcode])) == 0); } _PyExecutorObject *executor = make_executor_from_uops(buffer, &dependencies); if (executor == NULL) { return -1; } OPT_HIST(Py_SIZE(executor), optimized_trace_length_hist); *exec_ptr = executor; return 1; } static void uop_opt_dealloc(PyObject *self) { PyObject_Free(self); } PyTypeObject _PyUOpOptimizer_Type = { PyVarObject_HEAD_INIT(&PyType_Type, 0) .tp_name = "uop_optimizer", .tp_basicsize = sizeof(_PyOptimizerObject), .tp_itemsize = 0, .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_DISALLOW_INSTANTIATION, .tp_dealloc = uop_opt_dealloc, }; PyObject * PyUnstable_Optimizer_NewUOpOptimizer(void) { _PyOptimizerObject *opt = PyObject_New(_PyOptimizerObject, &_PyUOpOptimizer_Type); if (opt == NULL) { return NULL; } opt->optimize = uop_optimize; opt->resume_threshold = OPTIMIZER_UNREACHABLE_THRESHOLD; // Need a few iterations to settle specializations, // and to ammortize the cost of optimization. opt->side_threshold = 16; opt->backedge_threshold = 16; return (PyObject *)opt; } static void counter_dealloc(_PyExecutorObject *self) { /* The optimizer is the operand of the second uop. */ PyObject *opt = (PyObject *)self->trace[1].operand; Py_DECREF(opt); uop_dealloc(self); } PyTypeObject _PyCounterExecutor_Type = { PyVarObject_HEAD_INIT(&PyType_Type, 0) .tp_name = "counting_executor", .tp_basicsize = offsetof(_PyExecutorObject, exits), .tp_itemsize = 1, .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_DISALLOW_INSTANTIATION | Py_TPFLAGS_HAVE_GC, .tp_dealloc = (destructor)counter_dealloc, .tp_methods = executor_methods, .tp_traverse = executor_traverse, .tp_clear = executor_clear, }; static int counter_optimize( _PyOptimizerObject* self, _PyInterpreterFrame *frame, _Py_CODEUNIT *instr, _PyExecutorObject **exec_ptr, int Py_UNUSED(curr_stackentries) ) { PyCodeObject *code = (PyCodeObject *)frame->f_executable; int oparg = instr->op.arg; while (instr->op.code == EXTENDED_ARG) { instr++; oparg = (oparg << 8) | instr->op.arg; } if (instr->op.code != JUMP_BACKWARD) { /* Counter optimizer can only handle backward edges */ return 0; } _Py_CODEUNIT *target = instr + 1 + _PyOpcode_Caches[JUMP_BACKWARD] - oparg; _PyUOpInstruction buffer[3] = { { .opcode = _LOAD_CONST_INLINE_BORROW, .operand = (uintptr_t)self }, { .opcode = _INTERNAL_INCREMENT_OPT_COUNTER }, { .opcode = _EXIT_TRACE, .target = (uint32_t)(target - _PyCode_CODE(code)) } }; _PyExecutorObject *executor = make_executor_from_uops(buffer, &EMPTY_FILTER); if (executor == NULL) { return -1; } Py_INCREF(self); Py_SET_TYPE(executor, &_PyCounterExecutor_Type); *exec_ptr = executor; return 1; } static PyObject * counter_get_counter(PyObject *self, PyObject *args) { return PyLong_FromLongLong(((_PyCounterOptimizerObject *)self)->count); } static PyMethodDef counter_optimizer_methods[] = { { "get_count", counter_get_counter, METH_NOARGS, NULL }, { NULL, NULL }, }; PyTypeObject _PyCounterOptimizer_Type = { PyVarObject_HEAD_INIT(&PyType_Type, 0) .tp_name = "Counter optimizer", .tp_basicsize = sizeof(_PyCounterOptimizerObject), .tp_itemsize = 0, .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_DISALLOW_INSTANTIATION, .tp_methods = counter_optimizer_methods, .tp_dealloc = (destructor)PyObject_Del, }; PyObject * PyUnstable_Optimizer_NewCounter(void) { _PyCounterOptimizerObject *opt = (_PyCounterOptimizerObject *)_PyObject_New(&_PyCounterOptimizer_Type); if (opt == NULL) { return NULL; } opt->base.optimize = counter_optimize; opt->base.resume_threshold = OPTIMIZER_UNREACHABLE_THRESHOLD; opt->base.side_threshold = OPTIMIZER_UNREACHABLE_THRESHOLD; opt->base.backedge_threshold = 0; opt->count = 0; return (PyObject *)opt; } /***************************************** * Executor management ****************************************/ /* We use a bloomfilter with k = 6, m = 256 * The choice of k and the following constants * could do with a more rigourous analysis, * but here is a simple analysis: * * We want to keep the false positive rate low. * For n = 5 (a trace depends on 5 objects), * we expect 30 bits set, giving a false positive * rate of (30/256)**6 == 2.5e-6 which is plenty * good enough. * * However with n = 10 we expect 60 bits set (worst case), * giving a false positive of (60/256)**6 == 0.0001 * * We choose k = 6, rather than a higher number as * it means the false positive rate grows slower for high n. * * n = 5, k = 6 => fp = 2.6e-6 * n = 5, k = 8 => fp = 3.5e-7 * n = 10, k = 6 => fp = 1.6e-4 * n = 10, k = 8 => fp = 0.9e-4 * n = 15, k = 6 => fp = 0.18% * n = 15, k = 8 => fp = 0.23% * n = 20, k = 6 => fp = 1.1% * n = 20, k = 8 => fp = 2.3% * * The above analysis assumes perfect hash functions, * but those don't exist, so the real false positive * rates may be worse. */ #define K 6 #define SEED 20221211 /* TO DO -- Use more modern hash functions with better distribution of bits */ static uint64_t address_to_hash(void *ptr) { assert(ptr != NULL); uint64_t uhash = SEED; uintptr_t addr = (uintptr_t)ptr; for (int i = 0; i < SIZEOF_VOID_P; i++) { uhash ^= addr & 255; uhash *= (uint64_t)_PyHASH_MULTIPLIER; addr >>= 8; } return uhash; } void _Py_BloomFilter_Init(_PyBloomFilter *bloom) { for (int i = 0; i < BLOOM_FILTER_WORDS; i++) { bloom->bits[i] = 0; } } /* We want K hash functions that each set 1 bit. * A hash function that sets 1 bit in M bits can be trivially * derived from a log2(M) bit hash function. * So we extract 8 (log2(256)) bits at a time from * the 64bit hash. */ void _Py_BloomFilter_Add(_PyBloomFilter *bloom, void *ptr) { uint64_t hash = address_to_hash(ptr); assert(K <= 8); for (int i = 0; i < K; i++) { uint8_t bits = hash & 255; bloom->bits[bits >> 5] |= (1 << (bits&31)); hash >>= 8; } } static bool bloom_filter_may_contain(_PyBloomFilter *bloom, _PyBloomFilter *hashes) { for (int i = 0; i < BLOOM_FILTER_WORDS; i++) { if ((bloom->bits[i] & hashes->bits[i]) != hashes->bits[i]) { return false; } } return true; } static void link_executor(_PyExecutorObject *executor) { PyInterpreterState *interp = _PyInterpreterState_GET(); _PyExecutorLinkListNode *links = &executor->vm_data.links; _PyExecutorObject *head = interp->executor_list_head; if (head == NULL) { interp->executor_list_head = executor; links->previous = NULL; links->next = NULL; } else { _PyExecutorObject *next = head->vm_data.links.next; links->previous = head; links->next = next; if (next != NULL) { next->vm_data.links.previous = executor; } head->vm_data.links.next = executor; } executor->vm_data.valid = true; /* executor_list_head must be first in list */ assert(interp->executor_list_head->vm_data.links.previous == NULL); } static void unlink_executor(_PyExecutorObject *executor) { _PyExecutorLinkListNode *links = &executor->vm_data.links; _PyExecutorObject *next = links->next; _PyExecutorObject *prev = links->previous; if (next != NULL) { next->vm_data.links.previous = prev; } if (prev != NULL) { prev->vm_data.links.next = next; } else { // prev == NULL implies that executor is the list head PyInterpreterState *interp = PyInterpreterState_Get(); assert(interp->executor_list_head == executor); interp->executor_list_head = next; } executor->vm_data.valid = false; } /* This must be called by optimizers before using the executor */ void _Py_ExecutorInit(_PyExecutorObject *executor, const _PyBloomFilter *dependency_set) { executor->vm_data.valid = true; for (int i = 0; i < BLOOM_FILTER_WORDS; i++) { executor->vm_data.bloom.bits[i] = dependency_set->bits[i]; } link_executor(executor); } /* This must be called by executors during dealloc */ void _Py_ExecutorClear(_PyExecutorObject *executor) { if (!executor->vm_data.valid) { return; } unlink_executor(executor); PyCodeObject *code = executor->vm_data.code; if (code == NULL) { return; } for (uint32_t i = 0; i < executor->exit_count; i++) { Py_DECREF(executor->exits[i].executor); executor->exits[i].executor = &COLD_EXITS[i]; executor->exits[i].temperature = INT16_MIN; } _Py_CODEUNIT *instruction = &_PyCode_CODE(code)[executor->vm_data.index]; assert(instruction->op.code == ENTER_EXECUTOR); int index = instruction->op.arg; assert(code->co_executors->executors[index] == executor); instruction->op.code = executor->vm_data.opcode; instruction->op.arg = executor->vm_data.oparg; executor->vm_data.code = NULL; Py_CLEAR(code->co_executors->executors[index]); } void _Py_Executor_DependsOn(_PyExecutorObject *executor, void *obj) { assert(executor->vm_data.valid); _Py_BloomFilter_Add(&executor->vm_data.bloom, obj); } /* Invalidate all executors that depend on `obj` * May cause other executors to be invalidated as well */ void _Py_Executors_InvalidateDependency(PyInterpreterState *interp, void *obj, int is_invalidation) { _PyBloomFilter obj_filter; _Py_BloomFilter_Init(&obj_filter); _Py_BloomFilter_Add(&obj_filter, obj); /* Walk the list of executors */ /* TO DO -- Use a tree to avoid traversing as many objects */ for (_PyExecutorObject *exec = interp->executor_list_head; exec != NULL;) { assert(exec->vm_data.valid); _PyExecutorObject *next = exec->vm_data.links.next; if (bloom_filter_may_contain(&exec->vm_data.bloom, &obj_filter)) { _Py_ExecutorClear(exec); if (is_invalidation) { OPT_STAT_INC(executors_invalidated); } } exec = next; } } /* Invalidate all executors */ void _Py_Executors_InvalidateAll(PyInterpreterState *interp, int is_invalidation) { while (interp->executor_list_head) { _PyExecutorObject *executor = interp->executor_list_head; if (executor->vm_data.code) { // Clear the entire code object so its co_executors array be freed: _PyCode_Clear_Executors(executor->vm_data.code); } else { _Py_ExecutorClear(executor); } if (is_invalidation) { OPT_STAT_INC(executors_invalidated); } } }