summaryrefslogtreecommitdiffstats
path: root/src/common/uint128.h
blob: 83560a9ceefcfd219142fbd79fb0f36808582449 (plain) (blame)
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
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.

#pragma once

#include <cstring>
#include <utility>

#ifdef _MSC_VER
#include <intrin.h>
#pragma intrinsic(__umulh)
#pragma intrinsic(_umul128)
#pragma intrinsic(_udiv128)
#else
#include <x86intrin.h>
#endif

#include "common/common_types.h"

namespace Common {

// This function multiplies 2 u64 values and divides it by a u64 value.
[[nodiscard]] static inline u64 MultiplyAndDivide64(u64 a, u64 b, u64 d) {
#ifdef _MSC_VER
    u128 r{};
    r[0] = _umul128(a, b, &r[1]);
    u64 remainder;
#if _MSC_VER < 1923
    return udiv128(r[1], r[0], d, &remainder);
#else
    return _udiv128(r[1], r[0], d, &remainder);
#endif
#else
    const u64 diva = a / d;
    const u64 moda = a % d;
    const u64 divb = b / d;
    const u64 modb = b % d;
    return diva * b + moda * divb + moda * modb / d;
#endif
}

// This function multiplies 2 u64 values and produces a u128 value;
[[nodiscard]] static inline u128 Multiply64Into128(u64 a, u64 b) {
    u128 result;
#ifdef _MSC_VER
    result[0] = _umul128(a, b, &result[1]);
#else
    unsigned __int128 tmp = a;
    tmp *= b;
    std::memcpy(&result, &tmp, sizeof(u128));
#endif
    return result;
}

[[nodiscard]] static inline u64 GetFixedPoint64Factor(u64 numerator, u64 divisor) {
#ifdef __SIZEOF_INT128__
    const auto base = static_cast<unsigned __int128>(numerator) << 64ULL;
    return static_cast<u64>(base / divisor);
#elif defined(_M_X64) || defined(_M_ARM64)
    std::array<u64, 2> r = {0, numerator};
    u64 remainder;
#if _MSC_VER < 1923
    return udiv128(r[1], r[0], divisor, &remainder);
#else
    return _udiv128(r[1], r[0], divisor, &remainder);
#endif
#else
    // This one is bit more inaccurate.
    return MultiplyAndDivide64(std::numeric_limits<u64>::max(), numerator, divisor);
#endif
}

[[nodiscard]] static inline u64 MultiplyHigh(u64 a, u64 b) {
#ifdef __SIZEOF_INT128__
    return (static_cast<unsigned __int128>(a) * static_cast<unsigned __int128>(b)) >> 64;
#elif defined(_M_X64) || defined(_M_ARM64)
    return __umulh(a, b); // MSVC
#else
    // Generic fallback
    const u64 a_lo = u32(a);
    const u64 a_hi = a >> 32;
    const u64 b_lo = u32(b);
    const u64 b_hi = b >> 32;

    const u64 a_x_b_hi = a_hi * b_hi;
    const u64 a_x_b_mid = a_hi * b_lo;
    const u64 b_x_a_mid = b_hi * a_lo;
    const u64 a_x_b_lo = a_lo * b_lo;

    const u64 carry_bit = (static_cast<u64>(static_cast<u32>(a_x_b_mid)) +
                           static_cast<u64>(static_cast<u32>(b_x_a_mid)) + (a_x_b_lo >> 32)) >>
                          32;

    const u64 multhi = a_x_b_hi + (a_x_b_mid >> 32) + (b_x_a_mid >> 32) + carry_bit;

    return multhi;
#endif
}

} // namespace Common