Bitwise operations using binary representation of integers

Definitions

• bitwise operations for PosNum and Num,
• SNum, a type that represents integers as a bit string with a sign bit at the end,
• arithmetic operations for SNum.

def PosNum.lor : PosNum → PosNum → PosNum

Bitwise "or" for PosNum.

Equations

• PosNum.lor PosNum.one (PosNum.bit0 q) = PosNum.bit1 q
• PosNum.lor PosNum.one x = x
• PosNum.lor (PosNum.bit0 p) PosNum.one = PosNum.bit1 p
• PosNum.lor x PosNum.one = x
• PosNum.lor (PosNum.bit0 p) (PosNum.bit0 q) = PosNum.bit0 (PosNum.lor p q)
• PosNum.lor (PosNum.bit0 p) (PosNum.bit1 q) = PosNum.bit1 (PosNum.lor p q)
• PosNum.lor (PosNum.bit1 p) (PosNum.bit0 q) = PosNum.bit1 (PosNum.lor p q)
• PosNum.lor (PosNum.bit1 p) (PosNum.bit1 q) = PosNum.bit1 (PosNum.lor p q)

def PosNum.land : PosNum → PosNum → Num

Bitwise "and" for PosNum.

Equations

• PosNum.land PosNum.one (PosNum.bit0 q) = 0
• PosNum.land PosNum.one x = 1
• PosNum.land (PosNum.bit0 p) PosNum.one = 0
• PosNum.land x PosNum.one = 1
• PosNum.land (PosNum.bit0 p) (PosNum.bit0 q) = Num.bit0 (PosNum.land p q)
• PosNum.land (PosNum.bit0 p) (PosNum.bit1 q) = Num.bit0 (PosNum.land p q)
• PosNum.land (PosNum.bit1 p) (PosNum.bit0 q) = Num.bit0 (PosNum.land p q)
• PosNum.land (PosNum.bit1 p) (PosNum.bit1 q) = Num.bit1 (PosNum.land p q)

def PosNum.ldiff : PosNum → PosNum → Num

Bitwise fun a b ↦ a && !b for PosNum. For example, ldiff 5 9 = 4:

 101
1001
----
 100

Equations

• PosNum.ldiff PosNum.one (PosNum.bit0 a) = 1
• PosNum.ldiff PosNum.one x = 0
• PosNum.ldiff (PosNum.bit0 p) PosNum.one = Num.pos (PosNum.bit0 p)
• PosNum.ldiff (PosNum.bit1 p) PosNum.one = Num.pos (PosNum.bit0 p)
• PosNum.ldiff (PosNum.bit0 p) (PosNum.bit0 q) = Num.bit0 (PosNum.ldiff p q)
• PosNum.ldiff (PosNum.bit0 p) (PosNum.bit1 q) = Num.bit0 (PosNum.ldiff p q)
• PosNum.ldiff (PosNum.bit1 p) (PosNum.bit0 q) = Num.bit1 (PosNum.ldiff p q)
• PosNum.ldiff (PosNum.bit1 p) (PosNum.bit1 q) = Num.bit0 (PosNum.ldiff p q)

def PosNum.lxor : PosNum → PosNum → Num

Bitwise "xor" for PosNum.

Equations

• PosNum.lxor PosNum.one PosNum.one = 0
• PosNum.lxor PosNum.one (PosNum.bit0 q) = Num.pos (PosNum.bit1 q)
• PosNum.lxor PosNum.one (PosNum.bit1 q) = Num.pos (PosNum.bit0 q)
• PosNum.lxor (PosNum.bit0 p) PosNum.one = Num.pos (PosNum.bit1 p)
• PosNum.lxor (PosNum.bit1 p) PosNum.one = Num.pos (PosNum.bit0 p)
• PosNum.lxor (PosNum.bit0 p) (PosNum.bit0 q) = Num.bit0 (PosNum.lxor p q)
• PosNum.lxor (PosNum.bit0 p) (PosNum.bit1 q) = Num.bit1 (PosNum.lxor p q)
• PosNum.lxor (PosNum.bit1 p) (PosNum.bit0 q) = Num.bit1 (PosNum.lxor p q)
• PosNum.lxor (PosNum.bit1 p) (PosNum.bit1 q) = Num.bit0 (PosNum.lxor p q)

def PosNum.testBit : PosNum → ℕ → Bool

a.testBit n is true iff the n-th bit (starting from the LSB) in the binary representation of a is active. If the size of a is less than n, this evaluates to false.

Equations

• PosNum.testBit PosNum.one 0 = true
• PosNum.testBit PosNum.one x = false
• PosNum.testBit (PosNum.bit0 a) 0 = false
• PosNum.testBit (PosNum.bit0 p) (Nat.succ n) = PosNum.testBit p n
• PosNum.testBit (PosNum.bit1 a) 0 = true
• PosNum.testBit (PosNum.bit1 p) (Nat.succ n) = PosNum.testBit p n

def PosNum.oneBits : PosNum → ℕ → List ℕ

n.oneBits 0 is the list of indices of active bits in the binary representation of n.

Equations

• PosNum.oneBits PosNum.one x = [x]
• PosNum.oneBits (PosNum.bit0 p) x = PosNum.oneBits p (x + 1)
• PosNum.oneBits (PosNum.bit1 p) x = x :: PosNum.oneBits p (x + 1)

def PosNum.shiftl : PosNum → ℕ → PosNum

Left-shift the binary representation of a PosNum.

Equations

• PosNum.shiftl x 0 = x
• PosNum.shiftl x (Nat.succ n) = PosNum.shiftl (PosNum.bit0 x) n

theorem PosNum.shiftl_succ_eq_bit0_shiftl (p : PosNum) (n : ℕ) : PosNum.shiftl p (Nat.succ n) = PosNum.bit0 (PosNum.shiftl p n)

def PosNum.shiftr : PosNum → ℕ → Num

Right-shift the binary representation of a PosNum.

Equations

• PosNum.shiftr x 0 = Num.pos x
• PosNum.shiftr PosNum.one x = 0
• PosNum.shiftr (PosNum.bit0 p) (Nat.succ n) = PosNum.shiftr p n
• PosNum.shiftr (PosNum.bit1 p) (Nat.succ n) = PosNum.shiftr p n

def Num.lor : Num → Num → Num

Bitwise "or" for Num.

def Num.land : Num → Num → Num

Bitwise "and" for Num.

def Num.ldiff : Num → Num → Num

Bitwise fun a b ↦ a && !b for Num. For example, ldiff 5 9 = 4:

 101
1001
----
 100

def Num.lxor : Num → Num → Num

Bitwise "xor" for Num.

def Num.shiftl : Num → ℕ → Num

Left-shift the binary representation of a Num.

def Num.shiftr : Num → ℕ → Num

Right-shift the binary representation of a Num.

def Num.testBit : Num → ℕ → Bool

a.testBit n is true iff the n-th bit (starting from the LSB) in the binary representation of a is active. If the size of a is less than n, this evaluates to false.

def Num.oneBits : Num → List ℕ

n.oneBits is the list of indices of active bits in the binary representation of n.

inductive NzsNum : Type

• msb: Bool → NzsNum
• bit: Bool → NzsNum → NzsNum

  Add a bit at the end of a NzsNum.

This is a nonzero (and "non minus one") version of SNum. See the documentation of SNum for more details.

instance instDecidableEqNzsNum : DecidableEq NzsNum

inductive SNum : Type

• zero: Bool → SNum
• nz: NzsNum → SNum

Alternative representation of integers using a sign bit at the end. The convention on sign here is to have the argument to msb denote the sign of the MSB itself, with all higher bits set to the negation of this sign. The result is interpreted in two's complement.

13 = ..0001101(base 2) = nz (bit1 (bit0 (bit1 (msb true)))) -13 = ..1110011(base 2) = nz (bit1 (bit1 (bit0 (msb false))))

As with Num, a special case must be added for zero, which has no msb, but by two's complement symmetry there is a second special case for -1. Here the Bool field indicates the sign of the number.

0 = ..0000000(base 2) = zero false -1 = ..1111111(base 2) = zero true

instance instDecidableEqSNum : DecidableEq SNum

instance instCoeNzsNumSNum : Coe NzsNum SNum

instance instZeroSNum : Zero SNum

instance instOneNzsNum : One NzsNum

instance instOneSNum : One SNum

instance instInhabitedNzsNum : Inhabited NzsNum

instance instInhabitedSNum : Inhabited SNum

The SNum representation uses a bit string, essentially a list of 0 (false) and 1 (true) bits, and the negation of the MSB is sign-extended to all higher bits.

def NzsNum.«term_::_» : Lean.TrailingParserDescr

Add a bit at the end of a NzsNum.

def NzsNum.sign : NzsNum → Bool

Sign of a NzsNum.

Equations

• NzsNum.sign (NzsNum.msb b) = !b
• NzsNum.sign (NzsNum.bit a p) = NzsNum.sign p

@[match_pattern]

def NzsNum.not : NzsNum → NzsNum

Bitwise not for NzsNum.

Equations

• NzsNum.not (NzsNum.msb b) = NzsNum.msb (decide ¬b = true)
• NzsNum.not (NzsNum.bit a p) = NzsNum.bit (decide ¬a = true) (NzsNum.not p)

def NzsNum.«term~_» : Lean.ParserDescr

Bitwise not for NzsNum.

def NzsNum.bit0 : NzsNum → NzsNum

Add an inactive bit at the end of a NzsNum. This mimics PosNum.bit0.

def NzsNum.bit1 : NzsNum → NzsNum

Add an active bit at the end of a NzsNum. This mimics PosNum.bit1.

def NzsNum.head : NzsNum → Bool

The head of a NzsNum is the boolean value of its LSB.

def NzsNum.tail : NzsNum → SNum

The tail of a NzsNum is the SNum obtained by removing the LSB. Edge cases: tail 1 = 0 and tail (-2) = -1.

def SNum.sign : SNum → Bool

Sign of a SNum.

@[match_pattern]

def SNum.not : SNum → SNum

Bitwise not for SNum.

def SNum.«term~_» : Lean.ParserDescr

Bitwise not for SNum.

@[match_pattern]

def SNum.bit : Bool → SNum → SNum

Add a bit at the end of a SNum. This mimics NzsNum.bit.

def SNum.«term_::_» : Lean.TrailingParserDescr

Add a bit at the end of a SNum. This mimics NzsNum.bit.

def SNum.bit0 : SNum → SNum

Add an inactive bit at the end of a SNum. This mimics ZNum.bit0.

def SNum.bit1 : SNum → SNum

Add an active bit at the end of a SNum. This mimics ZNum.bit1.

theorem SNum.bit_zero (b : Bool) : SNum.bit b (SNum.zero b) = SNum.zero b

theorem SNum.bit_one (b : Bool) : SNum.bit b (SNum.zero (decide ¬b = true)) = SNum.nz (NzsNum.msb b)

def NzsNum.drec' {C : SNum → Sort u_1} (z : (b : Bool) → C (SNum.zero b)) (s : (b : Bool) → (p : SNum) → C p → C (SNum.bit b p)) (p : NzsNum) : C (SNum.nz p)

A dependent induction principle for NzsNum, with base cases 0 : SNum and (-1) : SNum.

Equations

• NzsNum.drec' z s (NzsNum.msb b) = Eq.mpr (_ : C (SNum.nz (NzsNum.msb b)) = C (SNum.bit b (SNum.zero (decide ¬b = true)))) (s b (SNum.zero (decide ¬b = true)) (z (decide ¬b = true)))
• NzsNum.drec' z s (NzsNum.bit a p) = s a (SNum.nz p) (NzsNum.drec' z s p)

def SNum.head : SNum → Bool

The head of a SNum is the boolean value of its LSB.

def SNum.tail : SNum → SNum

The tail of a SNum is obtained by removing the LSB. Edge cases: tail 1 = 0, tail (-2) = -1, tail 0 = 0 and tail (-1) = -1.

def SNum.drec' {C : SNum → Sort u_1} (z : (b : Bool) → C (SNum.zero b)) (s : (b : Bool) → (p : SNum) → C p → C (SNum.bit b p)) (p : SNum) : C p

A dependent induction principle for SNum which avoids relying on NzsNum.

def SNum.rec' {α : Sort u_1} (z : Bool → α) (s : Bool → SNum → α → α) : SNum → α

An induction principle for SNum which avoids relying on NzsNum.

def SNum.testBit : ℕ → SNum → Bool

SNum.testBit n a is true iff the n-th bit (starting from the LSB) of a is active. If the size of a is less than n, this evaluates to false.

Equations

• SNum.testBit 0 x = SNum.head x
• SNum.testBit (Nat.succ n) x = SNum.testBit n (SNum.tail x)

def SNum.succ : SNum → SNum

The successor of a SNum (i.e. the operation adding one).

def SNum.pred : SNum → SNum

The predecessor of a SNum (i.e. the operation of removing one).

def SNum.neg (n : SNum) : SNum

The opposite of a SNum.

instance SNum.instNegSNum : Neg SNum

def SNum.czAdd : Bool → Bool → SNum → SNum

SNum.czAdd a b n is n + a - b (where a and b should be read as either 0 or 1). This is useful to implement the carry system in cAdd.

def SNum.bits : SNum → (n : ℕ) → Vector Bool n

a.bits n is the vector of the n first bits of a (starting from the LSB).

Equations

• SNum.bits x 0 = Vector.nil
• SNum.bits x (Nat.succ n) = SNum.head x ::ᵥ SNum.bits (SNum.tail x) n

def SNum.cAdd : SNum → SNum → Bool → SNum

SNum.cAdd n m a is n + m + a (where a should be read as either 0 or 1). a represents a carry bit.

def SNum.add (a : SNum) (b : SNum) : SNum

Add two SNums.

instance SNum.instAddSNum : Add SNum

def SNum.sub (a : SNum) (b : SNum) : SNum

Subtract two SNums.

instance SNum.instSubSNum : Sub SNum

def SNum.mul (a : SNum) : SNum → SNum

Multiply two SNums.

instance SNum.instMulSNum : Mul SNum