local Numeric = {} local BIT_SET_31 = 0x80000000 local BIT_SET_32 = 0x100000000 local K_ZERO, K_ONE, K_BIT_SET_26 local bit_lshift = bit32.lshift local bit_rshift = bit32.rshift local bit_arshift = bit32.arshift local bit_and = bit32.band local bit_or = bit32.bor local bit_xor = bit32.bxor local bit_not = bit32.bnot local bit_replace = bit32.replace local math_ceil = math.ceil local math_floor = math.floor local math_log = math.log local math_max = math.max local math_pow = math.pow local from_u32, into_u32, from_u64, into_u64 local num_add, num_subtract, num_multiply, num_divide_unsigned, num_negate local num_is_negative, num_is_zero, num_is_equal, num_is_less_unsigned, num_is_greater_unsigned -- TODO: Eventually support Vector3 function Numeric.from_u32(data_1, data_2) return { data_1, data_2 } end function Numeric.into_u32(data) return data[1], data[2] end function Numeric.from_u64(value) if value < 0 then return num_negate(from_u64(-value)) else return from_u32(value % BIT_SET_32, math_floor(value / BIT_SET_32)) end end function Numeric.into_u64(value) local data_1, data_2 = into_u32(value) return data_1 + data_2 * BIT_SET_32 end function Numeric.add(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) local data_1 = data_l_1 + data_r_1 local data_2 = data_l_2 + data_r_2 if data_1 >= BIT_SET_32 then data_1 = data_1 - BIT_SET_32 data_2 = data_2 + 1 end if data_2 >= BIT_SET_32 then data_2 = data_2 - BIT_SET_32 end return from_u32(data_1, data_2) end function Numeric.subtract(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) local data_1 = data_l_1 - data_r_1 local data_2 = data_l_2 - data_r_2 if data_1 < 0 then data_1 = data_1 + BIT_SET_32 data_2 = data_2 - 1 end if data_2 < 0 then data_2 = data_2 + BIT_SET_32 end return from_u32(data_1, data_2) end local function set_absolute(lhs, rhs) local has_negative = false if num_is_negative(lhs) then lhs = num_negate(lhs) has_negative = not has_negative end if num_is_negative(rhs) then rhs = num_negate(rhs) has_negative = not has_negative end return has_negative, lhs, rhs end function Numeric.multiply(lhs, rhs) if num_is_zero(lhs) or num_is_zero(rhs) then return K_ZERO end local has_negative has_negative, lhs, rhs = set_absolute(lhs, rhs) -- If both longs are small, use float multiplication if num_is_less_unsigned(lhs, K_BIT_SET_26) and num_is_less_unsigned(rhs, K_BIT_SET_26) then local data_l_1, _ = into_u32(lhs) local data_r_1, _ = into_u32(rhs) local result = from_u64(data_l_1 * data_r_1) if has_negative then result = num_negate(result) end return result end -- Divide each long into 4 chunks of 16 bits, and then add up 4x4 products. -- We can skip products that would overflow. local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) local a48 = bit_rshift(data_l_2, 16) local a32 = bit_and(data_l_2, 0xFFFF) local a16 = bit_rshift(data_l_1, 16) local a00 = bit_and(data_l_1, 0xFFFF) local b48 = bit_rshift(data_r_2, 16) local b32 = bit_and(data_r_2, 0xFFFF) local b16 = bit_rshift(data_r_1, 16) local b00 = bit_and(data_r_1, 0xFFFF) local c00 = a00 * b00 local c16 = bit_rshift(c00, 16) c00 = bit_and(c00, 0xFFFF) c16 = c16 + a16 * b00 local c32 = bit_rshift(c16, 16) c16 = bit_and(c16, 0xFFFF) c16 = c16 + a00 * b16 c32 = c32 + bit_rshift(c16, 16) c16 = bit_and(c16, 0xFFFF) c32 = c32 + a32 * b00 local c48 = bit_rshift(c32, 16) c32 = bit_and(c32, 0xFFFF) c32 = c32 + a16 * b16 c48 = c48 + bit_rshift(c32, 16) c32 = bit_and(c32, 0xFFFF) c32 = c32 + a00 * b32 c48 = c48 + bit_rshift(c32, 16) c32 = bit_and(c32, 0xFFFF) c48 = c48 + a48 * b00 + a32 * b16 + a16 * b32 + a00 * b48 c48 = bit_and(c48, 0xFFFF) local data_1 = bit_replace(c00, c16, 16, 16) local data_2 = bit_replace(c32, c48, 16, 16) local result = from_u32(data_1, data_2) if has_negative then result = num_negate(result) end return result end local function get_approx_delta(rem, rhs) local approx = math_max(1, math_floor(rem / rhs)) local log = math_ceil(math_log(approx, 2)) local delta = log <= 48 and 1 or math_pow(2, log - 48) return approx, delta end function Numeric.divide_unsigned(lhs, rhs) if num_is_zero(rhs) then error("division by zero") elseif num_is_zero(lhs) then return 0 end local rhs_number = into_u64(rhs) local rem = lhs local res = K_ZERO while num_is_greater_unsigned(rem, rhs) or num_is_equal(rem, rhs) do local res_approx, delta = get_approx_delta(into_u64(rem), rhs_number) local res_temp = from_u64(res_approx) local rem_temp = num_multiply(res_temp, rhs) while num_is_negative(rem_temp) or num_is_greater_unsigned(rem_temp, rem) do res_approx = res_approx - delta res_temp = from_u64(res_approx) rem_temp = num_multiply(res_temp, rhs) end if num_is_zero(res_temp) then res_temp = K_ONE end res = num_add(res, res_temp) rem = num_subtract(rem, rem_temp) end return res end function Numeric.divide_signed(lhs, rhs) local has_negative has_negative, lhs, rhs = set_absolute(lhs, rhs) local result = num_divide_unsigned(lhs, rhs) if has_negative then result = num_negate(result) end return result end function Numeric.negate(value) return num_add(bit_not(value), K_ONE) end function Numeric.bit_and(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) return from_u32(bit_and(data_l_1, data_r_1), bit_and(data_l_2, data_r_2)) end function Numeric.bit_not(value) local data_1, data_2 = into_u32(value) return from_u32(bit_not(data_1), bit_not(data_2)) end function Numeric.bit_or(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) return from_u32(bit_or(data_l_1, data_r_1), bit_or(data_l_2, data_r_2)) end function Numeric.bit_xor(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) return from_u32(bit_xor(data_l_1, data_r_1), bit_xor(data_l_2, data_r_2)) end function Numeric.shift_left(lhs, rhs) local count = into_u64(rhs) if count < 32 then local data_l_1, data_l_2 = into_u32(lhs) local data_1 = bit_lshift(data_l_1, count) local data_2 = bit_or(bit_lshift(data_l_2, count), bit_rshift(data_l_1, 32 - count)) return from_u32(data_1, data_2) elseif count == 32 then local data_l_1, _ = into_u32(lhs) return from_u32(0, data_l_1) else local data_l_1, _ = into_u32(lhs) return from_u32(0, bit_lshift(data_l_1, count - 32)) end end function Numeric.shift_right_unsigned(lhs, rhs) local count = into_u64(rhs) if count < 32 then local data_l_1, data_l_2 = into_u32(lhs) local data_1 = bit_or(bit_rshift(data_l_1, count), bit_lshift(data_l_2, 32 - count)) local data_2 = bit_rshift(data_l_2, count) return from_u32(data_1, data_2) elseif count == 32 then local _, data_l_2 = into_u32(lhs) return from_u32(data_l_2, 0) else local _, data_l_2 = into_u32(lhs) return from_u32(bit_rshift(data_l_2, count - 32), 0) end end function Numeric.shift_right_signed(lhs, rhs) local count = into_u64(rhs) if count < 32 then local data_l_1, data_l_2 = into_u32(lhs) local data_1 = bit_or(bit_rshift(data_l_1, count), bit_lshift(data_l_2, 32 - count)) local data_2 = bit_arshift(data_l_2, count) return from_u32(data_1, data_2) else local _, data_l_2 = into_u32(lhs) local data_1 = bit_arshift(data_l_2, count - 32) local data_2 = data_l_2 > BIT_SET_31 and BIT_SET_32 - 1 or 0 return from_u32(data_1, data_2) end end function Numeric.is_negative(value) local _, data_2 = into_u32(value) return data_2 > BIT_SET_31 end function Numeric.is_zero(value) local data_1, data_2 = into_u32(value) return data_1 == 0 and data_2 == 0 end function Numeric.is_equal(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) return data_l_1 == data_r_1 and data_l_2 == data_r_2 end function Numeric.is_less_unsigned(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) return data_l_2 < data_r_2 or (data_l_2 == data_r_2 and data_l_1 < data_r_1) end function Numeric.is_greater_unsigned(lhs, rhs) local data_l_1, data_l_2 = into_u32(lhs) local data_r_1, data_r_2 = into_u32(rhs) return data_l_2 > data_r_2 or (data_l_2 == data_r_2 and data_l_1 > data_r_1) end function Numeric.is_less_signed(lhs, rhs) local neg_a = num_is_negative(lhs) local neg_b = num_is_negative(rhs) if neg_a and not neg_b then return true elseif not neg_a and neg_b then return false else return num_is_negative(num_subtract(lhs, rhs)) end end function Numeric.is_greater_signed(lhs, rhs) local neg_a = num_is_negative(lhs) local neg_b = num_is_negative(rhs) if neg_a and not neg_b then return false elseif not neg_a and neg_b then return true else return num_is_negative(num_subtract(rhs, lhs)) end end from_u32 = Numeric.from_u32 into_u32 = Numeric.into_u32 from_u64 = Numeric.from_u64 into_u64 = Numeric.into_u64 num_add = Numeric.add num_subtract = Numeric.subtract num_multiply = Numeric.multiply num_divide_unsigned = Numeric.divide_unsigned num_negate = Numeric.negate num_is_negative = Numeric.is_negative num_is_zero = Numeric.is_zero num_is_equal = Numeric.is_equal num_is_less_unsigned = Numeric.is_less_unsigned num_is_greater_unsigned = Numeric.is_greater_unsigned K_ZERO = from_u64(0) K_ONE = from_u64(1) K_BIT_SET_26 = from_u64(0x4000000) Numeric.K_ZERO = K_ZERO Numeric.K_ONE = K_ONE return Numeric