`<cstdlib>`

From the <cstdlib> header we support a function and a structure that are analogous to std::div. These are of particular importance with this library since, unlike a hardware type, the compiler cannot fold a division followed by a modulo of the same operands into one operation. The division algorithm naturally produces both, so it costs nearly nothing above a division or a modulo to get both results.

#include <boost/int256/cstdlib.hpp>

Structures

namespace boost {
namespace int256 {

struct u256div_t
{
    uint256 quot;
    uint256 rem;
};

struct i256div_t
{
    int256 quot;
    int256 rem;
};

} // namespace int256
} // namespace boost

div

Using the structures defined above, the div function computes both quotient and remainder simultaneously:

namespace boost {
namespace int256 {

BOOST_INT256_HOST_DEVICE constexpr u256div_t div(const uint256& x, const uint256& y) noexcept;

BOOST_INT256_HOST_DEVICE constexpr i256div_t div(const int256& x, const int256& y) noexcept;

} // namespace int256
} // namespace boost

For any non-zero divisor the quot and rem values are the same as if you performed the division and the modulo separately, so quot * y + rem == x always holds; for u256div_t, rem < y as well. For the signed overload, quot’s sign is `sign(x) != sign(y) and rem takes the sign of the dividend x, exactly matching operator/ and operator%, and BOOST_INT256_INT256_MIN / -1 is the same defined wrap (quot == INT256_MIN, rem == 0) that those operators define.

Division by zero is undefined behavior, exactly as it is for operator/ and operator% and for the built-in integer types. The library performs no zero-divisor check, and a zero divisor seen during constant evaluation is a hard compile-time error.

See the div example for a program that uses it, including a divisor wide enough to take the general division path, and the signed sign combinations.

Edit this Page