1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166

use int::Int;

#[inline]
fn mulx_u8(x: u8, y: u8) -> (u8, u8) {
    let result: u16 = (x as u16) * (y as u16);
    let hi = (result >> 8) as u8;
    (result as u8, hi)
}


#[inline]
fn mulx_u16(x: u16, y: u16) -> (u16, u16) {
    let result: u32 = (x as u32) * (y as u32);
    let hi = (result >> 16) as u16;
    (result as u16, hi)
}

#[inline]
fn mulx_u32(x: u32, y: u32) -> (u32, u32) {
    let result: u64 = (x as u64) * (y as u64);
    let hi = (result >> 32) as u32;
    (result as u32, hi)
}

#[inline]
fn mulx_u64(x: u64, y: u64) -> (u64, u64) {
    #[cfg(RUSTC_IS_NIGHTLY)]
    {
        let result: u128 = (x as u128) * (y as u128);
        let hi = (result >> 64) as u64;
        (result as u64, hi)
    }
    #[cfg(not(RUSTC_IS_NIGHTLY))]
    {
        let u1 = x & 0xffffffff;
        let v1 = y & 0xffffffff;
        let t = u1 * v1;
        let w3 = t & 0xffffffff;
        let k = t >> 32;

        let x = x >> 32;
        let t1 = (x * v1) + k;
        let k1 = t1 & 0xffffffff;
        let w1 = t1 >> 32;

        let y1 = y >> 32;
        let t2 = (u1 * y1) + k1;
        let k2 = t2 >> 32;

        let hi = (x * y1) + w1 + k2;
        let lo = (t2 << 32) + w3;
        (lo, hi)
    }
}

/// Unsigned multiply without affecting flags.
///
/// Unsigned multiplication of `x` with `y` returning a pair `(lo, hi)` with
/// the low half and the high half of the result.
///
/// # Assembly Instructions
///
/// - [`MULX`](http://www.felixcloutier.com/x86/MULX.html):
///   - Description: Unsigned multiply without affecting flags.
///   - Architecture: x86.
///   - Instruction set: BMI2.
///   - Registers: 32/64 bit.
///
/// # Example
///
/// ```
/// use bitintr::x86::bmi2::*;
///
/// { // 8-bit
///   let a: u8 = 128;
///   let b: u8 = 128;
///   let (lo, hi): (u8, u8) = mulx(a, b);
///   // result = 16384 = 0b0100_0000_0000_0000u16
///   //                    ^~hi~~~~~ ^~lo~~~~~
///   assert_eq!(lo, 0b0000_0000);
///   assert_eq!(hi, 0b0100_0000);
/// }
/// { // 16-bit
///   let a: u16 = 65_500;
///   let b: u16 = 65_500;
///   let (lo, hi): (u16, u16)  = a.mulx(b);
///   // result = 4290250000 = 0b1111_1111_1011_1000_0000_0101_0001_0000u32
///   //                         ^~hi~~~~~~~~~~~~~~~ ^~lo~~~~~~~~~~~~~~~
///   assert_eq!(lo, 0b0000_0101_0001_0000);
///   assert_eq!(hi, 0b1111_1111_1011_1000);
/// }
/// { // 32-bit
///   let a: u32 = 4_294_967_200;
///   let b: u32 = 2;
///   let (lo, hi): (u32, u32)  = mulx(a, b);
///   // result = 8589934400
///   //        = 0b0001_1111_1111_1111_1111_1111_1111_0100_0000u64
///   //            ^~hi ^~lo~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
///   assert_eq!(lo, 0b1111_1111_1111_1111_1111_1111_0100_0000u32);
///   assert_eq!(hi, 0b0001u32);
/// }
/// { // 64-bit
///   let a: u64 = 9_223_372_036_854_775_800;
///   let b: u64 = 100;
///   let (lo, hi): (u64, u64)  = mulx(a, b);
///   // result = 922337203685477580000
///   //        = 0b00110001_11111111_11111111_11111111_11111111_11111111_11111111_11111100_11100000u128
///   //            ^~hi~~~~ ^~lo~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
///   assert_eq!(lo, 0b11111111_11111111_11111111_11111111_11111111_11111111_11111100_11100000u64);
///   assert_eq!(hi, 0b00110001u64);
/// }
/// { // 8-bit
///   let a: i8 = 128u8 as i8;
///   let b: i8 = 128u8 as i8;
///   let (lo, hi): (i8, i8) = mulx(a, b);
///   // result = _____ = 0b0100_0000_0000_0000u16
///   //                    ^~hi~~~~~ ^~lo~~~~~
///   assert_eq!(lo, 0b0000_0000);
///   assert_eq!(hi, 0b0100_0000);
/// }
/// { // 16-bit
///   let a: i16 = 65_500u16 as i16;
///   let b: i16 = 65_500u16 as i16;
///   let (lo, hi): (i16, i16)  = mulx(a, b);
///   // result = 4290250000 = 0b1111_1111_1011_1000_0000_0101_0001_0000u32
///   //                         ^~hi~~~~~~~~~~~~~~~ ^~lo~~~~~~~~~~~~~~~
///   assert_eq!(lo, 0b0000_0101_0001_0000u16 as i16);
///   assert_eq!(hi, 0b1111_1111_1011_1000u16 as i16);
/// }
/// ```
#[inline]
pub fn mulx<T: Int>(x: T, y: T) -> (T, T) {

    match T::bit_size().to_u8() {
        8 => {
            let (x, y) = mulx_u8(x.to_u8(), y.to_u8());
            (T::from_u8(x), T::from_u8(y))
        }
        16 => {
            let (x, y) = mulx_u16(x.to_u16(), y.to_u16());
            (T::from_u16(x), T::from_u16(y))
        }
        32 => {
            let (x, y) = mulx_u32(x.to_u32(), y.to_u32());
            (T::from_u32(x), T::from_u32(y))
        }
        64 => {
            let (x, y) = mulx_u64(x.to_u64(), y.to_u64());
            (T::from_u64(x), T::from_u64(y))
        }
        _ => unreachable!(),
    }
}

/// Method version of [`mulx`](fn.mulx.html).
pub trait MULX: Sized {
    #[inline]
    fn mulx(self, Self) -> (Self, Self);
}

impl<T: Int> MULX for T {
    #[inline]
    fn mulx(self, x: Self) -> (Self, Self) {
        mulx(self, x)
    }
}