| # Copyright 2019 Google LLC |
| # |
| # Licensed under the Apache License, Version 2.0 (the "License"); |
| # you may not use this file except in compliance with the License. |
| # You may obtain a copy of the License at |
| # |
| # https://www.apache.org/licenses/LICENSE-2.0 |
| # |
| # Unless required by applicable law or agreed to in writing, software |
| # distributed under the License is distributed on an "AS IS" BASIS, |
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| # See the License for the specific language governing permissions and |
| # limitations under the License. |
| |
| """Functions for proving mathematical properties of expressions.""" |
| |
| import math |
| import fractions |
| import operator |
| |
| from compiler.util import ir_data |
| from compiler.util import ir_data_utils |
| from compiler.util import ir_util |
| from compiler.util import traverse_ir |
| |
| |
| # Create a local alias for math.gcd with a fallback to fractions.gcd if it is |
| # not available. This can be dropped if pre-3.5 Python support is dropped. |
| if hasattr(math, 'gcd'): |
| _math_gcd = math.gcd |
| else: |
| _math_gcd = fractions.gcd |
| |
| |
| def compute_constraints_of_expression(expression, ir): |
| """Adds appropriate bounding constraints to the given expression.""" |
| if ir_util.is_constant_type(expression.type): |
| return |
| expression_variety = expression.WhichOneof("expression") |
| if expression_variety == "constant": |
| _compute_constant_value_of_constant(expression) |
| elif expression_variety == "constant_reference": |
| _compute_constant_value_of_constant_reference(expression, ir) |
| elif expression_variety == "function": |
| _compute_constraints_of_function(expression, ir) |
| elif expression_variety == "field_reference": |
| _compute_constraints_of_field_reference(expression, ir) |
| elif expression_variety == "builtin_reference": |
| _compute_constraints_of_builtin_value(expression) |
| elif expression_variety == "boolean_constant": |
| _compute_constant_value_of_boolean_constant(expression) |
| else: |
| assert False, "Unknown expression variety {!r}".format(expression_variety) |
| if expression.type.WhichOneof("type") == "integer": |
| _assert_integer_constraints(expression) |
| |
| |
| def _compute_constant_value_of_constant(expression): |
| value = expression.constant.value |
| expression.type.integer.modular_value = value |
| expression.type.integer.minimum_value = value |
| expression.type.integer.maximum_value = value |
| expression.type.integer.modulus = "infinity" |
| |
| |
| def _compute_constant_value_of_constant_reference(expression, ir): |
| referred_object = ir_util.find_object( |
| expression.constant_reference.canonical_name, ir) |
| expression = ir_data_utils.builder(expression) |
| if isinstance(referred_object, ir_data.EnumValue): |
| compute_constraints_of_expression(referred_object.value, ir) |
| assert ir_util.is_constant(referred_object.value) |
| new_value = str(ir_util.constant_value(referred_object.value)) |
| expression.type.enumeration.value = new_value |
| elif isinstance(referred_object, ir_data.Field): |
| assert ir_util.field_is_virtual(referred_object), ( |
| "Non-virtual non-enum-value constant reference should have been caught " |
| "in type_check.py") |
| compute_constraints_of_expression(referred_object.read_transform, ir) |
| expression.type.CopyFrom(referred_object.read_transform.type) |
| else: |
| assert False, "Unexpected constant reference type." |
| |
| |
| def _compute_constraints_of_function(expression, ir): |
| """Computes the known constraints of the result of a function.""" |
| for arg in expression.function.args: |
| compute_constraints_of_expression(arg, ir) |
| op = expression.function.function |
| if op in (ir_data.FunctionMapping.ADDITION, ir_data.FunctionMapping.SUBTRACTION): |
| _compute_constraints_of_additive_operator(expression) |
| elif op == ir_data.FunctionMapping.MULTIPLICATION: |
| _compute_constraints_of_multiplicative_operator(expression) |
| elif op in (ir_data.FunctionMapping.EQUALITY, ir_data.FunctionMapping.INEQUALITY, |
| ir_data.FunctionMapping.LESS, ir_data.FunctionMapping.LESS_OR_EQUAL, |
| ir_data.FunctionMapping.GREATER, ir_data.FunctionMapping.GREATER_OR_EQUAL, |
| ir_data.FunctionMapping.AND, ir_data.FunctionMapping.OR): |
| _compute_constant_value_of_comparison_operator(expression) |
| elif op == ir_data.FunctionMapping.CHOICE: |
| _compute_constraints_of_choice_operator(expression) |
| elif op == ir_data.FunctionMapping.MAXIMUM: |
| _compute_constraints_of_maximum_function(expression) |
| elif op == ir_data.FunctionMapping.PRESENCE: |
| _compute_constraints_of_existence_function(expression, ir) |
| elif op in (ir_data.FunctionMapping.UPPER_BOUND, ir_data.FunctionMapping.LOWER_BOUND): |
| _compute_constraints_of_bound_function(expression) |
| else: |
| assert False, "Unknown operator {!r}".format(op) |
| |
| |
| def _compute_constraints_of_existence_function(expression, ir): |
| """Computes the constraints of a $has(field) expression.""" |
| field_path = expression.function.args[0].field_reference.path[-1] |
| field = ir_util.find_object(field_path, ir) |
| compute_constraints_of_expression(field.existence_condition, ir) |
| ir_data_utils.builder(expression).type.CopyFrom(field.existence_condition.type) |
| |
| |
| def _compute_constraints_of_field_reference(expression, ir): |
| """Computes the constraints of a reference to a structure's field.""" |
| field_path = expression.field_reference.path[-1] |
| field = ir_util.find_object(field_path, ir) |
| if isinstance(field, ir_data.Field) and ir_util.field_is_virtual(field): |
| # References to virtual fields should have the virtual field's constraints |
| # copied over. |
| compute_constraints_of_expression(field.read_transform, ir) |
| ir_data_utils.builder(expression).type.CopyFrom(field.read_transform.type) |
| return |
| # Non-virtual non-integer fields do not (yet) have constraints. |
| if expression.type.WhichOneof("type") == "integer": |
| # TODO(bolms): These lines will need to change when support is added for |
| # fixed-point types. |
| expression.type.integer.modulus = "1" |
| expression.type.integer.modular_value = "0" |
| type_definition = ir_util.find_parent_object(field_path, ir) |
| if isinstance(field, ir_data.Field): |
| referrent_type = field.type |
| else: |
| referrent_type = field.physical_type_alias |
| if referrent_type.HasField("size_in_bits"): |
| type_size = ir_util.constant_value(referrent_type.size_in_bits) |
| else: |
| field_size = ir_util.constant_value(field.location.size) |
| if field_size is None: |
| type_size = None |
| else: |
| type_size = field_size * type_definition.addressable_unit |
| assert referrent_type.HasField("atomic_type"), field |
| assert not referrent_type.atomic_type.reference.canonical_name.module_file |
| _set_integer_constraints_from_physical_type( |
| expression, referrent_type, type_size) |
| |
| |
| def _set_integer_constraints_from_physical_type( |
| expression, physical_type, type_size): |
| """Copies the integer constraints of an expression from a physical type.""" |
| # SCAFFOLDING HACK: In order to keep changelists manageable, this hardcodes |
| # the ranges for all of the Emboss Prelude integer types. This would break |
| # any user-defined `external` integer types, but that feature isn't fully |
| # implemented in the C++ backend, so it doesn't matter for now. |
| # |
| # Adding the attribute(s) for integer bounds will require new operators: |
| # integer/flooring division, remainder, and exponentiation (2**N, 10**N). |
| # |
| # (Technically, there are a few sets of operators that would work: for |
| # example, just the choice operator `?:` is sufficient, but very ugly. |
| # Bitwise AND, bitshift, and exponentiation would also work, but `10**($bits |
| # >> 2) * 2**($bits & 0b11) - 1` isn't quite as clear as `10**($bits // 4) * |
| # 2**($bits % 4) - 1`, in my (bolms@) opinion.) |
| # |
| # TODO(bolms): Add a scheme for defining integer bounds on user-defined |
| # external types. |
| if type_size is None: |
| # If the type_size is unknown, then we can't actually say anything about the |
| # minimum and maximum values of the type. For UInt, Int, and Bcd, an error |
| # will be thrown during the constraints check stage. |
| expression.type.integer.minimum_value = "-infinity" |
| expression.type.integer.maximum_value = "infinity" |
| return |
| name = tuple(physical_type.atomic_type.reference.canonical_name.object_path) |
| if name == ("UInt",): |
| expression.type.integer.minimum_value = "0" |
| expression.type.integer.maximum_value = str(2**type_size - 1) |
| elif name == ("Int",): |
| expression.type.integer.minimum_value = str(-(2**(type_size - 1))) |
| expression.type.integer.maximum_value = str(2**(type_size - 1) - 1) |
| elif name == ("Bcd",): |
| expression.type.integer.minimum_value = "0" |
| expression.type.integer.maximum_value = str( |
| 10**(type_size // 4) * 2**(type_size % 4) - 1) |
| else: |
| assert False, "Unknown integral type " + ".".join(name) |
| |
| |
| def _compute_constraints_of_parameter(parameter): |
| if parameter.type.WhichOneof("type") == "integer": |
| type_size = ir_util.constant_value( |
| parameter.physical_type_alias.size_in_bits) |
| _set_integer_constraints_from_physical_type( |
| parameter, parameter.physical_type_alias, type_size) |
| |
| |
| def _compute_constraints_of_builtin_value(expression): |
| """Computes the constraints of a builtin (like $static_size_in_bits).""" |
| name = expression.builtin_reference.canonical_name.object_path[0] |
| if name == "$static_size_in_bits": |
| expression.type.integer.modulus = "1" |
| expression.type.integer.modular_value = "0" |
| expression.type.integer.minimum_value = "0" |
| # The maximum theoretically-supported size of something is 2**64 bytes, |
| # which is 2**64 * 8 bits. |
| # |
| # Really, $static_size_in_bits is only valid in expressions that have to be |
| # evaluated at compile time anyway, so it doesn't really matter if the |
| # bounds are excessive. |
| expression.type.integer.maximum_value = "infinity" |
| elif name == "$is_statically_sized": |
| # No bounds on a boolean variable. |
| pass |
| elif name == "$logical_value": |
| # $logical_value is the placeholder used in inferred write-through |
| # transformations. |
| # |
| # Only integers (currently) have "real" write-through transformations, but |
| # fields that would otherwise be straight aliases, but which have a |
| # [requires] attribute, are elevated to write-through fields, so that the |
| # [requires] clause can be checked in Write, CouldWriteValue, TryToWrite, |
| # Read, and Ok. |
| if expression.type.WhichOneof("type") == "integer": |
| assert expression.type.integer.modulus |
| assert expression.type.integer.modular_value |
| assert expression.type.integer.minimum_value |
| assert expression.type.integer.maximum_value |
| elif expression.type.WhichOneof("type") == "enumeration": |
| assert expression.type.enumeration.name |
| elif expression.type.WhichOneof("type") == "boolean": |
| pass |
| else: |
| assert False, "Unexpected type for $logical_value" |
| else: |
| assert False, "Unknown builtin " + name |
| |
| |
| def _compute_constant_value_of_boolean_constant(expression): |
| expression.type.boolean.value = expression.boolean_constant.value |
| |
| |
| def _add(a, b): |
| """Adds a and b, where a and b are ints, "infinity", or "-infinity".""" |
| if a in ("infinity", "-infinity"): |
| a, b = b, a |
| if b == "infinity": |
| assert a != "-infinity" |
| return "infinity" |
| if b == "-infinity": |
| assert a != "infinity" |
| return "-infinity" |
| return int(a) + int(b) |
| |
| |
| def _sub(a, b): |
| """Subtracts b from a, where a and b are ints, "infinity", or "-infinity".""" |
| if b == "infinity": |
| return _add(a, "-infinity") |
| if b == "-infinity": |
| return _add(a, "infinity") |
| return _add(a, -int(b)) |
| |
| |
| def _sign(a): |
| """Returns 1 if a > 0, 0 if a == 0, and -1 if a < 0.""" |
| if a == "infinity": |
| return 1 |
| if a == "-infinity": |
| return -1 |
| if int(a) > 0: |
| return 1 |
| if int(a) < 0: |
| return -1 |
| return 0 |
| |
| |
| def _mul(a, b): |
| """Multiplies a and b, where a and b are ints, "infinity", or "-infinity".""" |
| if _is_infinite(a): |
| a, b = b, a |
| if _is_infinite(b): |
| sign = _sign(a) * _sign(b) |
| if sign > 0: |
| return "infinity" |
| if sign < 0: |
| return "-infinity" |
| return 0 |
| return int(a) * int(b) |
| |
| |
| def _is_infinite(a): |
| return a in ("infinity", "-infinity") |
| |
| |
| def _max(a): |
| """Returns max of a, where elements are ints, "infinity", or "-infinity".""" |
| if any(n == "infinity" for n in a): |
| return "infinity" |
| if all(n == "-infinity" for n in a): |
| return "-infinity" |
| return max(int(n) for n in a if not _is_infinite(n)) |
| |
| |
| def _min(a): |
| """Returns min of a, where elements are ints, "infinity", or "-infinity".""" |
| if any(n == "-infinity" for n in a): |
| return "-infinity" |
| if all(n == "infinity" for n in a): |
| return "infinity" |
| return min(int(n) for n in a if not _is_infinite(n)) |
| |
| |
| def _compute_constraints_of_additive_operator(expression): |
| """Computes the modular value of an additive expression.""" |
| funcs = { |
| ir_data.FunctionMapping.ADDITION: _add, |
| ir_data.FunctionMapping.SUBTRACTION: _sub, |
| } |
| func = funcs[expression.function.function] |
| args = expression.function.args |
| for arg in args: |
| assert arg.type.integer.modular_value, str(expression) |
| left, right = args |
| unadjusted_modular_value = func(left.type.integer.modular_value, |
| right.type.integer.modular_value) |
| new_modulus = _greatest_common_divisor(left.type.integer.modulus, |
| right.type.integer.modulus) |
| expression.type.integer.modulus = str(new_modulus) |
| if new_modulus == "infinity": |
| expression.type.integer.modular_value = str(unadjusted_modular_value) |
| else: |
| expression.type.integer.modular_value = str(unadjusted_modular_value % |
| new_modulus) |
| lmax = left.type.integer.maximum_value |
| lmin = left.type.integer.minimum_value |
| if expression.function.function == ir_data.FunctionMapping.SUBTRACTION: |
| rmax = right.type.integer.minimum_value |
| rmin = right.type.integer.maximum_value |
| else: |
| rmax = right.type.integer.maximum_value |
| rmin = right.type.integer.minimum_value |
| expression.type.integer.minimum_value = str(func(lmin, rmin)) |
| expression.type.integer.maximum_value = str(func(lmax, rmax)) |
| |
| |
| def _compute_constraints_of_multiplicative_operator(expression): |
| """Computes the modular value of a multiplicative expression.""" |
| bounds = [arg.type.integer for arg in expression.function.args] |
| |
| # The minimum and maximum values can come from any of the four pairings of |
| # (left min, left max) with (right min, right max), depending on the signs and |
| # magnitudes of the minima and maxima. E.g.: |
| # |
| # max = left max * right max: [ 2, 3] * [ 2, 3] |
| # max = left min * right min: [-3, -2] * [-3, -2] |
| # max = left max * right min: [-3, -2] * [ 2, 3] |
| # max = left min * right max: [ 2, 3] * [-3, -2] |
| # max = left max * right max: [-2, 3] * [-2, 3] |
| # max = left min * right min: [-3, 2] * [-3, 2] |
| # |
| # For uncorrelated multiplication, the minimum and maximum will always come |
| # from multiplying one extreme by another: if x is nonzero, then |
| # |
| # (y + e) * x > y * x || (y - e) * x > y * x |
| # |
| # for arbitrary nonzero e, so the extrema can only occur when we either cannot |
| # add or cannot subtract e. |
| # |
| # Correlated multiplication (e.g., `x * x`) can have tighter bounds, but |
| # Emboss is not currently trying to be that smart. |
| lmin, lmax = bounds[0].minimum_value, bounds[0].maximum_value |
| rmin, rmax = bounds[1].minimum_value, bounds[1].maximum_value |
| extrema = [_mul(lmax, rmax), _mul(lmin, rmax), # |
| _mul(lmax, rmin), _mul(lmin, rmin)] |
| expression.type.integer.minimum_value = str(_min(extrema)) |
| expression.type.integer.maximum_value = str(_max(extrema)) |
| |
| if all(bound.modulus == "infinity" for bound in bounds): |
| # If both sides are constant, the result is constant. |
| expression.type.integer.modulus = "infinity" |
| expression.type.integer.modular_value = str(int(bounds[0].modular_value) * |
| int(bounds[1].modular_value)) |
| return |
| |
| if any(bound.modulus == "infinity" for bound in bounds): |
| # If one side is constant and the other is not, then the non-constant |
| # modulus and modular_value can both be multiplied by the constant. E.g., |
| # if `a` is congruent to 3 mod 5, then `4 * a` will be congruent to 12 mod |
| # 20: |
| # |
| # a = ... | 4 * a = ... | 4 * a mod 20 = ... |
| # 3 | 12 | 12 |
| # 8 | 32 | 12 |
| # 13 | 52 | 12 |
| # 18 | 72 | 12 |
| # 23 | 92 | 12 |
| # 28 | 112 | 12 |
| # 33 | 132 | 12 |
| # |
| # This is trivially shown by noting that the difference between consecutive |
| # possible values for `4 * a` always differ by 20. |
| if bounds[0].modulus == "infinity": |
| constant, variable = bounds |
| else: |
| variable, constant = bounds |
| if int(constant.modular_value) == 0: |
| # If the constant is 0, the result is 0, no matter what the variable side |
| # is. |
| expression.type.integer.modulus = "infinity" |
| expression.type.integer.modular_value = "0" |
| return |
| new_modulus = int(variable.modulus) * abs(int(constant.modular_value)) |
| expression.type.integer.modulus = str(new_modulus) |
| # The `% new_modulus` will force the `modular_value` to be positive, even |
| # when `constant.modular_value` is negative. |
| expression.type.integer.modular_value = str( |
| int(variable.modular_value) * int(constant.modular_value) % new_modulus) |
| return |
| |
| # If neither side is constant, then the result is more complex. Full proof is |
| # available in g3doc/modular_congruence_multiplication_proof.md |
| # |
| # Essentially, if: |
| # |
| # l == _ * l_mod + l_mv |
| # r == _ * r_mod + r_mv |
| # |
| # Then we find l_mod0 and r_mod0 in: |
| # |
| # l == (_ * l_mod_nz + l_mv_nz) * l_mod0 |
| # r == (_ * r_mod_nz + r_mv_nz) * r_mod0 |
| # |
| # And finally conclude: |
| # |
| # l * r == _ * GCD(l_mod_nz, r_mod_nz) * l_mod0 * r_mod0 + l_mv * r_mv |
| product_of_zero_congruence_moduli = 1 |
| product_of_modular_values = 1 |
| nonzero_congruence_moduli = [] |
| for bound in bounds: |
| zero_congruence_modulus = _greatest_common_divisor(bound.modulus, |
| bound.modular_value) |
| assert int(bound.modulus) % zero_congruence_modulus == 0 |
| product_of_zero_congruence_moduli *= zero_congruence_modulus |
| product_of_modular_values *= int(bound.modular_value) |
| nonzero_congruence_moduli.append(int(bound.modulus) // |
| zero_congruence_modulus) |
| shared_nonzero_congruence_modulus = _greatest_common_divisor( |
| nonzero_congruence_moduli[0], nonzero_congruence_moduli[1]) |
| final_modulus = (shared_nonzero_congruence_modulus * |
| product_of_zero_congruence_moduli) |
| expression.type.integer.modulus = str(final_modulus) |
| expression.type.integer.modular_value = str(product_of_modular_values % |
| final_modulus) |
| |
| |
| def _assert_integer_constraints(expression): |
| """Asserts that the integer bounds of expression are self-consistent. |
| |
| Asserts that `minimum_value` and `maximum_value` are congruent to |
| `modular_value` modulo `modulus`. |
| |
| If `modulus` is "infinity", asserts that `minimum_value`, `maximum_value`, and |
| `modular_value` are all equal. |
| |
| If `minimum_value` is equal to `maximum_value`, asserts that `modular_value` |
| is equal to both, and that `modulus` is "infinity". |
| |
| Arguments: |
| expression: an expression with type.integer |
| |
| Returns: |
| None |
| """ |
| bounds = expression.type.integer |
| if bounds.modulus == "infinity": |
| assert bounds.minimum_value == bounds.modular_value |
| assert bounds.maximum_value == bounds.modular_value |
| return |
| modulus = int(bounds.modulus) |
| assert modulus > 0 |
| if bounds.minimum_value != "-infinity": |
| assert int(bounds.minimum_value) % modulus == int(bounds.modular_value) |
| if bounds.maximum_value != "infinity": |
| assert int(bounds.maximum_value) % modulus == int(bounds.modular_value) |
| if bounds.minimum_value == bounds.maximum_value: |
| # TODO(bolms): I believe there are situations using the not-yet-implemented |
| # integer division operator that would trigger these asserts, so they should |
| # be turned into assignments (with corresponding tests) when implementing |
| # division. |
| assert bounds.modular_value == bounds.minimum_value |
| assert bounds.modulus == "infinity" |
| if bounds.minimum_value != "-infinity" and bounds.maximum_value != "infinity": |
| assert int(bounds.minimum_value) <= int(bounds.maximum_value) |
| |
| |
| def _compute_constant_value_of_comparison_operator(expression): |
| """Computes the constant value, if any, of a comparison operator.""" |
| args = expression.function.args |
| if all(ir_util.is_constant(arg) for arg in args): |
| functions = { |
| ir_data.FunctionMapping.EQUALITY: operator.eq, |
| ir_data.FunctionMapping.INEQUALITY: operator.ne, |
| ir_data.FunctionMapping.LESS: operator.lt, |
| ir_data.FunctionMapping.LESS_OR_EQUAL: operator.le, |
| ir_data.FunctionMapping.GREATER: operator.gt, |
| ir_data.FunctionMapping.GREATER_OR_EQUAL: operator.ge, |
| ir_data.FunctionMapping.AND: operator.and_, |
| ir_data.FunctionMapping.OR: operator.or_, |
| } |
| func = functions[expression.function.function] |
| expression.type.boolean.value = func( |
| *[ir_util.constant_value(arg) for arg in args]) |
| |
| |
| def _compute_constraints_of_bound_function(expression): |
| """Computes the constraints of $upper_bound or $lower_bound.""" |
| if expression.function.function == ir_data.FunctionMapping.UPPER_BOUND: |
| value = expression.function.args[0].type.integer.maximum_value |
| elif expression.function.function == ir_data.FunctionMapping.LOWER_BOUND: |
| value = expression.function.args[0].type.integer.minimum_value |
| else: |
| assert False, "Non-bound function" |
| expression.type.integer.minimum_value = value |
| expression.type.integer.maximum_value = value |
| expression.type.integer.modular_value = value |
| expression.type.integer.modulus = "infinity" |
| |
| |
| def _compute_constraints_of_maximum_function(expression): |
| """Computes the constraints of the $max function.""" |
| assert expression.type.WhichOneof("type") == "integer" |
| args = expression.function.args |
| assert args[0].type.WhichOneof("type") == "integer" |
| # The minimum value of the result occurs when every argument takes its minimum |
| # value, which means that the minimum result is the maximum-of-minimums. |
| expression.type.integer.minimum_value = str(_max( |
| [arg.type.integer.minimum_value for arg in args])) |
| # The maximum result is the maximum-of-maximums. |
| expression.type.integer.maximum_value = str(_max( |
| [arg.type.integer.maximum_value for arg in args])) |
| # If the expression is dominated by a constant factor, then the result is |
| # constant. I (bolms@) believe this is the only case where |
| # _compute_constraints_of_maximum_function might violate the assertions in |
| # _assert_integer_constraints. |
| if (expression.type.integer.minimum_value == |
| expression.type.integer.maximum_value): |
| expression.type.integer.modular_value = ( |
| expression.type.integer.minimum_value) |
| expression.type.integer.modulus = "infinity" |
| return |
| result_modulus = args[0].type.integer.modulus |
| result_modular_value = args[0].type.integer.modular_value |
| # The result of $max(a, b) could be either a or b, which means that the result |
| # of $max(a, b) uses the _shared_modular_value() of a and b, just like the |
| # choice operator '?:'. |
| # |
| # This also takes advantage of the fact that $max(a, b, c, d, ...) is |
| # equivalent to $max(a, $max(b, $max(c, $max(d, ...)))), so it is valid to |
| # call _shared_modular_value() in a loop. |
| for arg in args[1:]: |
| # TODO(bolms): I think the bounds could be tigher in some cases where |
| # arg.maximum_value is less than the new expression.minimum_value, and |
| # in some very specific cases where arg.maximum_value is greater than the |
| # new expression.minimum_value, but arg.maximum_value - arg.modulus is less |
| # than expression.minimum_value. |
| result_modulus, result_modular_value = _shared_modular_value( |
| (result_modulus, result_modular_value), |
| (arg.type.integer.modulus, arg.type.integer.modular_value)) |
| expression.type.integer.modulus = str(result_modulus) |
| expression.type.integer.modular_value = str(result_modular_value) |
| |
| |
| def _shared_modular_value(left, right): |
| """Returns the shared modulus and modular value of left and right. |
| |
| Arguments: |
| left: A tuple of (modulus, modular value) |
| right: A tuple of (modulus, modular value) |
| |
| Returns: |
| A tuple of (modulus, modular_value) such that: |
| |
| left.modulus % result.modulus == 0 |
| right.modulus % result.modulus == 0 |
| left.modular_value % result.modulus = result.modular_value |
| right.modular_value % result.modulus = result.modular_value |
| |
| That is, the result.modulus and result.modular_value will be compatible |
| with, but (possibly) less restrictive than both left.(modulus, |
| modular_value) and right.(modulus, modular_value). |
| """ |
| left_modulus, left_modular_value = left |
| right_modulus, right_modular_value = right |
| # The combined modulus is gcd(gcd(left_modulus, right_modulus), |
| # left_modular_value - right_modular_value). |
| # |
| # The inner gcd normalizes the left_modulus and right_modulus, but can leave |
| # incompatible modular_values. The outer gcd finds a modulus to which both |
| # modular_values are congruent. Some examples: |
| # |
| # left | right | res |
| # --------------+----------------+-------------------- |
| # l % 12 == 7 | r % 20 == 15 | res % 4 == 3 |
| # l == 35 | r % 20 == 15 | res % 20 == 15 |
| # l % 24 == 15 | r % 12 == 7 | res % 4 == 3 |
| # l % 20 == 15 | r % 20 == 10 | res % 5 == 0 |
| # l % 20 == 16 | r % 20 == 11 | res % 5 == 1 |
| # l == 10 | r == 7 | res % 3 == 1 |
| # l == 4 | r == 4 | res == 4 |
| # |
| # The cases where one side or the other are constant are handled |
| # automatically by the fact that _greatest_common_divisor("infinity", x) |
| # is x. |
| common_modulus = _greatest_common_divisor(left_modulus, right_modulus) |
| new_modulus = _greatest_common_divisor( |
| common_modulus, abs(int(left_modular_value) - int(right_modular_value))) |
| if new_modulus == "infinity": |
| # The only way for the new_modulus to come out as "infinity" *should* be |
| # if both if_true and if_false have the same constant value. |
| assert left_modular_value == right_modular_value |
| assert left_modulus == right_modulus == "infinity" |
| return new_modulus, left_modular_value |
| else: |
| assert (int(left_modular_value) % new_modulus == |
| int(right_modular_value) % new_modulus) |
| return new_modulus, int(left_modular_value) % new_modulus |
| |
| |
| def _compute_constraints_of_choice_operator(expression): |
| """Computes the constraints of a choice operation '?:'.""" |
| condition, if_true, if_false = ir_data_utils.reader(expression).function.args |
| expression = ir_data_utils.builder(expression) |
| if condition.type.boolean.HasField("value"): |
| # The generated expressions for $size_in_bits and $size_in_bytes look like |
| # |
| # $max((field1_existence_condition ? field1_start + field1_size : 0), |
| # (field2_existence_condition ? field2_start + field2_size : 0), |
| # (field3_existence_condition ? field3_start + field3_size : 0), |
| # ...) |
| # |
| # Since most existence_conditions are just "true", it is important to select |
| # the tighter bounds in those cases -- otherwise, only zero-length |
| # structures could have a constant $size_in_bits or $size_in_bytes. |
| side = if_true if condition.type.boolean.value else if_false |
| expression.type.CopyFrom(side.type) |
| return |
| # The type.integer minimum_value/maximum_value bounding code is needed since |
| # constraints.check_constraints() will complain if minimum and maximum are not |
| # set correctly. I'm (bolms@) not sure if the modulus/modular_value pulls its |
| # weight, but for completeness I've left it in. |
| if if_true.type.WhichOneof("type") == "integer": |
| # The minimum value of the choice is the minimum value of either side, and |
| # the maximum is the maximum value of either side. |
| expression.type.integer.minimum_value = str(_min([ |
| if_true.type.integer.minimum_value, |
| if_false.type.integer.minimum_value])) |
| expression.type.integer.maximum_value = str(_max([ |
| if_true.type.integer.maximum_value, |
| if_false.type.integer.maximum_value])) |
| new_modulus, new_modular_value = _shared_modular_value( |
| (if_true.type.integer.modulus, if_true.type.integer.modular_value), |
| (if_false.type.integer.modulus, if_false.type.integer.modular_value)) |
| expression.type.integer.modulus = str(new_modulus) |
| expression.type.integer.modular_value = str(new_modular_value) |
| else: |
| assert if_true.type.WhichOneof("type") in ("boolean", "enumeration"), ( |
| "Unknown type {} for expression".format( |
| if_true.type.WhichOneof("type"))) |
| |
| |
| def _greatest_common_divisor(a, b): |
| """Returns the greatest common divisor of a and b. |
| |
| Arguments: |
| a: an integer, a stringified integer, or the string "infinity" |
| b: an integer, a stringified integer, or the string "infinity" |
| |
| Returns: |
| Conceptually, "infinity" is treated as the product of all integers. |
| |
| If both a and b are 0, returns "infinity". |
| |
| Otherwise, if either a or b are "infinity", and the other is 0, returns |
| "infinity". |
| |
| Otherwise, if either a or b are "infinity", returns the other. |
| |
| Otherwise, returns the greatest common divisor of a and b. |
| """ |
| if a != "infinity": a = int(a) |
| if b != "infinity": b = int(b) |
| assert a == "infinity" or a >= 0 |
| assert b == "infinity" or b >= 0 |
| if a == b == 0: return "infinity" |
| # GCD(0, x) is always x, so it's safe to shortcut when a == 0 or b == 0. |
| if a == 0: return b |
| if b == 0: return a |
| if a == "infinity": return b |
| if b == "infinity": return a |
| return _math_gcd(a, b) |
| |
| |
| def compute_constants(ir): |
| """Computes constant values for all expressions in ir. |
| |
| compute_constants calculates all constant values and adds them to the type |
| information for each expression and subexpression. |
| |
| Arguments: |
| ir: an IR on which to compute constants |
| |
| Returns: |
| A (possibly empty) list of errors. |
| """ |
| traverse_ir.fast_traverse_ir_top_down( |
| ir, [ir_data.Expression], compute_constraints_of_expression, |
| skip_descendants_of={ir_data.Expression}) |
| traverse_ir.fast_traverse_ir_top_down( |
| ir, [ir_data.RuntimeParameter], _compute_constraints_of_parameter, |
| skip_descendants_of={ir_data.Expression}) |
| return [] |