mul_128_256_shift

fn mul_128_256_shift(a: *const [2]u64, b: *const [4]u64, shift: u32, corr: u32) [4]u64

File

Code

fn mul_128_256_shift(a: *const [2]u64, b: *const [4]u64, shift: u32, corr: u32) [4]u64 {
    std.debug.assert(shift > 0);
    std.debug.assert(shift < 256);

    const b00 = @as(u128, a[0]) * b[0];
    const b01 = @as(u128, a[0]) * b[1];
    const b02 = @as(u128, a[0]) * b[2];
    const b03 = @as(u128, a[0]) * b[3];
    const b10 = @as(u128, a[1]) * b[0];
    const b11 = @as(u128, a[1]) * b[1];
    const b12 = @as(u128, a[1]) * b[2];
    const b13 = @as(u128, a[1]) * b[3];

    const s0 = b00;
    const s1 = b01 +% b10;
    const c1: u128 = @intFromBool(s1 < b01);
    const s2 = b02 +% b11;
    const c2: u128 = @intFromBool(s2 < b02);
    const s3 = b03 +% b12;
    const c3: u128 = @intFromBool(s3 < b03);

    const p0 = s0 +% (s1 << 64);
    const d0: u128 = @intFromBool(p0 < b00);
    const q1 = s2 +% (s1 >> 64) +% (s3 << 64);
    const d1: u128 = @intFromBool(q1 < s2);
    const p1 = q1 +% (c1 << 64) +% d0;
    const d2: u128 = @intFromBool(p1 < q1);
    const p2 = b13 +% (s3 >> 64) +% c2 +% (c3 << 64) +% d1 +% d2;

    var r0: u128 = undefined;
    var r1: u128 = undefined;
    if (shift < 128) {
        const cshift: u7 = @intCast(shift);
        const sshift: u7 = @intCast(128 - shift);
        r0 = corr +% ((p0 >> cshift) | (p1 << sshift));
        r1 = ((p1 >> cshift) | (p2 << sshift)) +% @intFromBool(r0 < corr);
    } else if (shift == 128) {
        r0 = corr +% p1;
        r1 = p2 +% @intFromBool(r0 < corr);
    } else {
        const ashift: u7 = @intCast(shift - 128);
        const sshift: u7 = @intCast(256 - shift);
        r0 = corr +% ((p1 >> ashift) | (p2 << sshift));
        r1 = (p2 >> ashift) +% @intFromBool(r0 < corr);
    }

    return .{ @truncate(r0), @truncate(r0 >> 64), @truncate(r1), @truncate(r1 >> 64) };
}