Theory Code_Target_Integer_Bit
chapter ‹Bit operations for target language integers›
theory Code_Target_Integer_Bit
imports
"HOL-Library.Word"
"Code_Int_Integer_Conversion"
"Word_Lib.Most_significant_bit"
"Word_Lib.Generic_set_bit"
"Word_Lib.Bit_Comprehension"
begin
text ‹TODO: separate›
lemmas [transfer_rule] =
identity_quotient
fun_quotient
Quotient_integer[folded integer.pcr_cr_eq]
lemma undefined_transfer:
assumes "Quotient R Abs Rep T"
shows "T (Rep undefined) undefined"
using assms unfolding Quotient_alt_def by blast
bundle undefined_transfer = undefined_transfer[transfer_rule]
section ‹More lemmas about @{typ integer}s›
context
includes integer.lifting
begin
lemma integer_of_nat_less_0_conv [simp]: "¬ integer_of_nat n < 0"
by transfer simp
lemma int_of_integer_pow: "int_of_integer (x ^ n) = int_of_integer x ^ n"
by transfer rule
lemma sub1_lt_0_iff [simp]: "Code_Numeral.sub n num.One < 0 ⟷ False"
by transfer (simp add: sub_negative)
lemma nat_of_integer_numeral [simp]: "nat_of_integer (numeral n) = numeral n"
by transfer simp
lemma nat_of_integer_sub1_conv_pred_numeral [simp]:
"nat_of_integer (Code_Numeral.sub n num.One) = pred_numeral n"
by transfer (simp only: pred_numeral_def int_nat_eq numeral_One int_minus flip: int_int_eq, simp)
lemma nat_of_integer_1 [simp]: "nat_of_integer 1 = 1"
by transfer simp
lemma dup_1 [simp]: "Code_Numeral.dup 1 = 2"
by transfer simp
end
section ‹Target language implementations›
text ‹
Unfortunately, this is not straightforward,
because these API functions have different signatures and preconditions on the parameters:
\begin{description}
\item[Standard ML] Shifts in IntInf are given as word, but not IntInf.
\item[Haskell] In the Data.Bits.Bits type class, shifts and bit indices are given as Int rather than Integer.
\end{description}
Additional constants take only parameters of type @{typ integer} rather than @{typ nat}
and check the preconditions as far as possible (e.g., being non-negative) in a portable way.
Manual implementations inside code\_printing perform the remaining range checks and convert
these @{typ integer}s into the right type.
For normalisation by evaluation, we derive custom code equations, because NBE
does not know these code\_printing serialisations and would otherwise loop.
›
code_printing code_module Integer_Bit ⇀ (SML)
‹structure Integer_Bit : sig
val test_bit : IntInf.int -> IntInf.int -> bool
val set_bit : IntInf.int -> IntInf.int -> bool -> IntInf.int
val shiftl : IntInf.int -> IntInf.int -> IntInf.int
val shiftr : IntInf.int -> IntInf.int -> IntInf.int
end = struct
val maxWord = IntInf.pow (2, Word.wordSize);
fun test_bit x n =
if n < maxWord then IntInf.andb (x, IntInf.<< (1, Word.fromLargeInt (IntInf.toLarge n))) <> 0
else raise (Fail ("Bit index too large: " ^ IntInf.toString n));
fun set_bit x n b =
if n < maxWord then
if b then IntInf.orb (x, IntInf.<< (1, Word.fromLargeInt (IntInf.toLarge n)))
else IntInf.andb (x, IntInf.notb (IntInf.<< (1, Word.fromLargeInt (IntInf.toLarge n))))
else raise (Fail ("Bit index too large: " ^ IntInf.toString n));
fun shiftl x n =
if n < maxWord then IntInf.<< (x, Word.fromLargeInt (IntInf.toLarge n))
else raise (Fail ("Shift operand too large: " ^ IntInf.toString n));
fun shiftr x n =
if n < maxWord then IntInf.~>> (x, Word.fromLargeInt (IntInf.toLarge n))
else raise (Fail ("Shift operand too large: " ^ IntInf.toString n));
end; (*struct Integer_Bit*)›
code_reserved SML Integer_Bit
code_printing code_module Integer_Bit ⇀ (OCaml)
‹module Integer_Bit : sig
val test_bit : Z.t -> Z.t -> bool
val shiftl : Z.t -> Z.t -> Z.t
val shiftr : Z.t -> Z.t -> Z.t
end = struct
(* We do not need an explicit range checks here,
because Big_int.int_of_big_int raises Failure
if the argument does not fit into an int. *)
let test_bit x n = Z.testbit x (Z.to_int n);;
let shiftl x n = Z.shift_left x (Z.to_int n);;
let shiftr x n = Z.shift_right x (Z.to_int n);;
end;; (*struct Integer_Bit*)›
code_reserved OCaml Integer_Bit
code_printing code_module Data_Bits ⇀ (Haskell)
‹
module Data_Bits where {
import qualified Data.Bits;
{-
The ...Bounded functions assume that the Integer argument for the shift
or bit index fits into an Int, is non-negative and (for types of fixed bit width)
less than bitSize
-}
infixl 7 .&.;
infixl 6 `xor`;
infixl 5 .|.;
(.&.) :: Data.Bits.Bits a => a -> a -> a;
(.&.) = (Data.Bits..&.);
xor :: Data.Bits.Bits a => a -> a -> a;
xor = Data.Bits.xor;
(.|.) :: Data.Bits.Bits a => a -> a -> a;
(.|.) = (Data.Bits..|.);
complement :: Data.Bits.Bits a => a -> a;
complement = Data.Bits.complement;
testBitUnbounded :: Data.Bits.Bits a => a -> Integer -> Bool;
testBitUnbounded x b
| b <= toInteger (Prelude.maxBound :: Int) = Data.Bits.testBit x (fromInteger b)
| otherwise = error ("Bit index too large: " ++ show b)
;
testBitBounded :: Data.Bits.Bits a => a -> Integer -> Bool;
testBitBounded x b = Data.Bits.testBit x (fromInteger b);
setBitUnbounded :: Data.Bits.Bits a => a -> Integer -> Bool -> a;
setBitUnbounded x n b
| n <= toInteger (Prelude.maxBound :: Int) =
if b then Data.Bits.setBit x (fromInteger n) else Data.Bits.clearBit x (fromInteger n)
| otherwise = error ("Bit index too large: " ++ show n)
;
setBitBounded :: Data.Bits.Bits a => a -> Integer -> Bool -> a;
setBitBounded x n True = Data.Bits.setBit x (fromInteger n);
setBitBounded x n False = Data.Bits.clearBit x (fromInteger n);
shiftlUnbounded :: Data.Bits.Bits a => a -> Integer -> a;
shiftlUnbounded x n
| n <= toInteger (Prelude.maxBound :: Int) = Data.Bits.shiftL x (fromInteger n)
| otherwise = error ("Shift operand too large: " ++ show n)
;
shiftlBounded :: Data.Bits.Bits a => a -> Integer -> a;
shiftlBounded x n = Data.Bits.shiftL x (fromInteger n);
shiftrUnbounded :: Data.Bits.Bits a => a -> Integer -> a;
shiftrUnbounded x n
| n <= toInteger (Prelude.maxBound :: Int) = Data.Bits.shiftR x (fromInteger n)
| otherwise = error ("Shift operand too large: " ++ show n)
;
shiftrBounded :: (Ord a, Data.Bits.Bits a) => a -> Integer -> a;
shiftrBounded x n = Data.Bits.shiftR x (fromInteger n);
}›
and
(Haskell_Quickcheck)
‹
module Data_Bits where {
import qualified Data.Bits;
{-
The functions assume that the Int argument for the shift or bit index is
non-negative and (for types of fixed bit width) less than bitSize
-}
infixl 7 .&.;
infixl 6 `xor`;
infixl 5 .|.;
(.&.) :: Data.Bits.Bits a => a -> a -> a;
(.&.) = (Data.Bits..&.);
xor :: Data.Bits.Bits a => a -> a -> a;
xor = Data.Bits.xor;
(.|.) :: Data.Bits.Bits a => a -> a -> a;
(.|.) = (Data.Bits..|.);
complement :: Data.Bits.Bits a => a -> a;
complement = Data.Bits.complement;
testBitUnbounded :: Data.Bits.Bits a => a -> Prelude.Int -> Bool;
testBitUnbounded = Data.Bits.testBit;
testBitBounded :: Data.Bits.Bits a => a -> Prelude.Int -> Bool;
testBitBounded = Data.Bits.testBit;
setBitUnbounded :: Data.Bits.Bits a => a -> Prelude.Int -> Bool -> a;
setBitUnbounded x n True = Data.Bits.setBit x n;
setBitUnbounded x n False = Data.Bits.clearBit x n;
setBitBounded :: Data.Bits.Bits a => a -> Prelude.Int -> Bool -> a;
setBitBounded x n True = Data.Bits.setBit x n;
setBitBounded x n False = Data.Bits.clearBit x n;
shiftlUnbounded :: Data.Bits.Bits a => a -> Prelude.Int -> a;
shiftlUnbounded = Data.Bits.shiftL;
shiftlBounded :: Data.Bits.Bits a => a -> Prelude.Int -> a;
shiftlBounded = Data.Bits.shiftL;
shiftrUnbounded :: Data.Bits.Bits a => a -> Prelude.Int -> a;
shiftrUnbounded = Data.Bits.shiftR;
shiftrBounded :: (Ord a, Data.Bits.Bits a) => a -> Prelude.Int -> a;
shiftrBounded = Data.Bits.shiftR;
}›
code_reserved Haskell Data_Bits
code_printing code_module Integer_Bit ⇀ (Scala)
‹object Integer_Bit {
def testBit(x: BigInt, n: BigInt) : Boolean =
n.isValidInt match {
case true => x.testBit(n.toInt)
case false => sys.error("Bit index too large: " + n.toString)
}
def setBit(x: BigInt, n: BigInt, b: Boolean) : BigInt =
n.isValidInt match {
case true if b => x.setBit(n.toInt)
case true => x.clearBit(n.toInt)
case false => sys.error("Bit index too large: " + n.toString)
}
def shiftl(x: BigInt, n: BigInt) : BigInt =
n.isValidInt match {
case true => x << n.toInt
case false => sys.error("Shift index too large: " + n.toString)
}
def shiftr(x: BigInt, n: BigInt) : BigInt =
n.isValidInt match {
case true => x << n.toInt
case false => sys.error("Shift index too large: " + n.toString)
}
} /* object Integer_Bit */›
code_printing
constant "Bit_Operations.and :: integer ⇒ integer ⇒ integer" ⇀
(SML) "IntInf.andb ((_),/ (_))" and
(OCaml) "Z.logand" and
(Haskell) "((Data'_Bits..&.) :: Integer -> Integer -> Integer)" and
(Haskell_Quickcheck) "((Data'_Bits..&.) :: Prelude.Int -> Prelude.Int -> Prelude.Int)" and
(Scala) infixl 3 "&"
| constant "Bit_Operations.or :: integer ⇒ integer ⇒ integer" ⇀
(SML) "IntInf.orb ((_),/ (_))" and
(OCaml) "Z.logor" and
(Haskell) "((Data'_Bits..|.) :: Integer -> Integer -> Integer)" and
(Haskell_Quickcheck) "((Data'_Bits..|.) :: Prelude.Int -> Prelude.Int -> Prelude.Int)" and
(Scala) infixl 1 "|"
| constant "Bit_Operations.xor :: integer ⇒ integer ⇒ integer" ⇀
(SML) "IntInf.xorb ((_),/ (_))" and
(OCaml) "Z.logxor" and
(Haskell) "(Data'_Bits.xor :: Integer -> Integer -> Integer)" and
(Haskell_Quickcheck) "(Data'_Bits.xor :: Prelude.Int -> Prelude.Int -> Prelude.Int)" and
(Scala) infixl 2 "^"
| constant "Bit_Operations.not :: integer ⇒ integer" ⇀
(SML) "IntInf.notb" and
(OCaml) "Z.lognot" and
(Haskell) "(Data'_Bits.complement :: Integer -> Integer)" and
(Haskell_Quickcheck) "(Data'_Bits.complement :: Prelude.Int -> Prelude.Int)" and
(Scala) "_.unary'_~"
definition integer_test_bit :: "integer ⇒ integer ⇒ bool"
where "integer_test_bit x n = (if n < 0 then undefined x n else bit x (nat_of_integer n))"
lemma integer_test_bit_code [code]:
"integer_test_bit x (Code_Numeral.Neg n) = undefined x (Code_Numeral.Neg n)"
"integer_test_bit 0 0 = False"
"integer_test_bit 0 (Code_Numeral.Pos n) = False"
"integer_test_bit (Code_Numeral.Pos num.One) 0 = True"
"integer_test_bit (Code_Numeral.Pos (num.Bit0 n)) 0 = False"
"integer_test_bit (Code_Numeral.Pos (num.Bit1 n)) 0 = True"
"integer_test_bit (Code_Numeral.Pos num.One) (Code_Numeral.Pos n') = False"
"integer_test_bit (Code_Numeral.Pos (num.Bit0 n)) (Code_Numeral.Pos n') =
integer_test_bit (Code_Numeral.Pos n) (Code_Numeral.sub n' num.One)"
"integer_test_bit (Code_Numeral.Pos (num.Bit1 n)) (Code_Numeral.Pos n') =
integer_test_bit (Code_Numeral.Pos n) (Code_Numeral.sub n' num.One)"
"integer_test_bit (Code_Numeral.Neg num.One) 0 = True"
"integer_test_bit (Code_Numeral.Neg (num.Bit0 n)) 0 = False"
"integer_test_bit (Code_Numeral.Neg (num.Bit1 n)) 0 = True"
"integer_test_bit (Code_Numeral.Neg num.One) (Code_Numeral.Pos n') = True"
"integer_test_bit (Code_Numeral.Neg (num.Bit0 n)) (Code_Numeral.Pos n') =
integer_test_bit (Code_Numeral.Neg n) (Code_Numeral.sub n' num.One)"
"integer_test_bit (Code_Numeral.Neg (num.Bit1 n)) (Code_Numeral.Pos n') =
integer_test_bit (Code_Numeral.Neg (n + num.One)) (Code_Numeral.sub n' num.One)"
by (simp_all add: integer_test_bit_def bit_integer_def bit_0 flip: bit_not_int_iff')
lemma bit_integer_code [code]:
"bit x n ⟷ integer_test_bit x (integer_of_nat n)"
by (simp add: integer_test_bit_def)
code_printing constant integer_test_bit ⇀
(SML) "Integer'_Bit.test'_bit" and
(OCaml) "Integer'_Bit.test'_bit" and
(Haskell) "(Data'_Bits.testBitUnbounded :: Integer -> Integer -> Bool)" and
(Haskell_Quickcheck) "(Data'_Bits.testBitUnbounded :: Prelude.Int -> Prelude.Int -> Bool)" and
(Scala) "Integer'_Bit.testBit"
context
includes integer.lifting and bit_operations_syntax
begin
lemma msb_integer_code [code]:
"msb (x :: integer) ⟷ x < 0"
by transfer (simp add: msb_int_def)
definition integer_set_bit :: "integer ⇒ integer ⇒ bool ⇒ integer"
where [code del]: "integer_set_bit x n b = (if n < 0 then undefined x n b else set_bit x (nat_of_integer n) b)"
lemma set_bit_integer_code [code]:
"set_bit x i b = integer_set_bit x (integer_of_nat i) b"
by(simp add: integer_set_bit_def)
lemma set_bit_integer_conv_masks:
fixes x :: integer shows
"set_bit x i b = (if b then x OR (push_bit i 1) else x AND NOT (push_bit i 1))"
by (transfer; rule bit_eqI) (simp add: bit_simps)
end
code_printing constant integer_set_bit ⇀
(SML) "Integer'_Bit.set'_bit" and
(Haskell) "(Data'_Bits.setBitUnbounded :: Integer -> Integer -> Bool -> Integer)" and
(Haskell_Quickcheck) "(Data'_Bits.setBitUnbounded :: Prelude.Int -> Prelude.Int -> Bool -> Prelude.Int)" and
(Scala) "Integer'_Bit.setBit"
text ‹
OCaml.Big\_int does not have a method for changing an individual bit, so we emulate that with masks.
We prefer an Isabelle implementation, because this then takes care of the signs for AND and OR.
›
context
includes bit_operations_syntax
begin
lemma integer_set_bit_code [code]:
"integer_set_bit x n b =
(if n < 0 then undefined x n b else
if b then x OR (push_bit (nat_of_integer n) 1)
else x AND NOT (push_bit (nat_of_integer n) 1))"
by (auto simp add: integer_set_bit_def not_less set_bit_eq set_bit_def unset_bit_def)
end
definition integer_shiftl :: "integer ⇒ integer ⇒ integer"
where [code del]: "integer_shiftl x n = (if n < 0 then undefined x n else push_bit (nat_of_integer n) x)"
declare [[code drop: ‹push_bit :: nat ⇒ integer ⇒ integer›]]
lemma shiftl_integer_code [code]:
fixes x :: integer shows
"push_bit n x = integer_shiftl x (integer_of_nat n)"
by(auto simp add: integer_shiftl_def)
context
includes integer.lifting
begin
lemma shiftl_integer_conv_mult_pow2:
fixes x :: integer shows
"push_bit n x = x * 2 ^ n"
by (fact push_bit_eq_mult)
lemma integer_shiftl_code [code]:
"integer_shiftl x (Code_Numeral.Neg n) = undefined x (Code_Numeral.Neg n)"
"integer_shiftl x 0 = x"
"integer_shiftl x (Code_Numeral.Pos n) = integer_shiftl (Code_Numeral.dup x) (Code_Numeral.sub n num.One)"
"integer_shiftl 0 (Code_Numeral.Pos n) = 0"
apply (simp_all add: integer_shiftl_def numeral_eq_Suc)
apply transfer
apply (simp add: ac_simps)
done
end
code_printing constant integer_shiftl ⇀
(SML) "Integer'_Bit.shiftl" and
(OCaml) "Integer'_Bit.shiftl" and
(Haskell) "(Data'_Bits.shiftlUnbounded :: Integer -> Integer -> Integer)" and
(Haskell_Quickcheck) "(Data'_Bits.shiftlUnbounded :: Prelude.Int -> Prelude.Int -> Prelude.Int)" and
(Scala) "Integer'_Bit.shiftl"
definition integer_shiftr :: "integer ⇒ integer ⇒ integer"
where [code del]: "integer_shiftr x n = (if n < 0 then undefined x n else drop_bit (nat_of_integer n) x)"
declare [[code drop: ‹drop_bit :: nat ⇒ integer ⇒ integer›]]
lemma shiftr_integer_conv_div_pow2:
includes integer.lifting fixes x :: integer shows
"drop_bit n x = x div 2 ^ n"
by (fact drop_bit_eq_div)
lemma shiftr_integer_code [code]:
fixes x :: integer shows
"drop_bit n x = integer_shiftr x (integer_of_nat n)"
by(auto simp add: integer_shiftr_def)
code_printing constant integer_shiftr ⇀
(SML) "Integer'_Bit.shiftr" and
(OCaml) "Integer'_Bit.shiftr" and
(Haskell) "(Data'_Bits.shiftrUnbounded :: Integer -> Integer -> Integer)" and
(Haskell_Quickcheck) "(Data'_Bits.shiftrUnbounded :: Prelude.Int -> Prelude.Int -> Prelude.Int)" and
(Scala) "Integer'_Bit.shiftr"
lemma integer_shiftr_code [code]:
includes integer.lifting
shows
"integer_shiftr x (Code_Numeral.Neg n) = undefined x (Code_Numeral.Neg n)"
"integer_shiftr x 0 = x"
"integer_shiftr 0 (Code_Numeral.Pos n) = 0"
"integer_shiftr (Code_Numeral.Pos num.One) (Code_Numeral.Pos n) = 0"
"integer_shiftr (Code_Numeral.Pos (num.Bit0 n')) (Code_Numeral.Pos n) =
integer_shiftr (Code_Numeral.Pos n') (Code_Numeral.sub n num.One)"
"integer_shiftr (Code_Numeral.Pos (num.Bit1 n')) (Code_Numeral.Pos n) =
integer_shiftr (Code_Numeral.Pos n') (Code_Numeral.sub n num.One)"
"integer_shiftr (Code_Numeral.Neg num.One) (Code_Numeral.Pos n) = -1"
"integer_shiftr (Code_Numeral.Neg (num.Bit0 n')) (Code_Numeral.Pos n) =
integer_shiftr (Code_Numeral.Neg n') (Code_Numeral.sub n num.One)"
"integer_shiftr (Code_Numeral.Neg (num.Bit1 n')) (Code_Numeral.Pos n) =
integer_shiftr (Code_Numeral.Neg (Num.inc n')) (Code_Numeral.sub n num.One)"
apply (simp_all add: integer_shiftr_def numeral_eq_Suc drop_bit_Suc)
apply transfer apply simp
apply transfer apply simp
apply transfer apply (simp add: add_One)
done
context
includes integer.lifting and bit_operations_syntax
begin
definition odd_integer :: ‹integer ⇒ bool›
where ‹odd_integer = odd›
lemma odd_integer_code [code]:
‹odd_integer i ⟷ i AND 1 ≠ 0›
by (simp add: odd_integer_def and_one_eq odd_iff_mod_2_eq_one)
lemma odd_integer_code_nbe [code nbe]:
‹odd_integer i ⟷ i mod 2 ≠ 0›
by (simp add: odd_integer_def odd_iff_mod_2_eq_one)
definition Bit_Cons_integer :: ‹bool ⇒ integer ⇒ integer›
where ‹Bit_Cons_integer b k = of_bool b + 2 * k›
lemma bit_Bit_Cons_integer_iff:
‹bit (Bit_Cons_integer b k) n ⟷ (if n = 0 then b else bit k (n - 1))›
by (simp add: Bit_Cons_integer_def bit_simps even_bit_succ_iff)
lemma Bit_Cons_integer_code [code]:
"Bit_Cons_integer False i = push_bit 1 i"
"Bit_Cons_integer True i = push_bit 1 i + 1"
by (simp_all add: Bit_Cons_integer_def)
lemma bitAND_integer_unfold [code]:
"x AND y =
(if x = 0 then 0
else if x = - 1 then y
else Bit_Cons_integer (odd_integer x ∧ odd_integer y) (drop_bit 1 x AND drop_bit 1 y))"
apply (rule bit_eqI)
apply (simp add: bit_simps bit_Bit_Cons_integer_iff odd_integer_def bit_0)
done
lemma bitOR_integer_unfold [code]:
"x OR y =
(if x = 0 then y
else if x = - 1 then - 1
else Bit_Cons_integer (odd_integer x ∨ odd_integer y) (drop_bit 1 x OR drop_bit 1 y))"
apply (rule bit_eqI)
apply (simp add: bit_simps bit_Bit_Cons_integer_iff odd_integer_def bit_0)
done
lemma bitXOR_integer_unfold [code]:
"x XOR y =
(if x = 0 then y
else if x = - 1 then NOT y
else Bit_Cons_integer (¬ odd_integer x ⟷ odd_integer y) (drop_bit 1 x XOR drop_bit 1 y))"
apply (rule bit_eqI)
apply (auto simp add: bit_simps bit_Bit_Cons_integer_iff odd_integer_def bit_0)
done
end
end