C++ 如何在 Visual Studio 上启用 __int128？

How to enable __int128 on Visual Studio?

提问by Amir Saniyan

When I type __int128in a C++ project in Visual Studio, the editor changes color of __int128to blue (like keyword).

当我__int128在 Visual Studio 中输入 C++ 项目时，编辑器将颜色更改__int128为蓝色（如关键字）。

But when I compile the source, the folowing error appears:

但是当我编译源代码时，出现以下错误：

error C4235: 
nonstandard extension used : '__int128' keyword not supported on this architecture

How can I enable __int128on Visual Studio?

如何__int128在 Visual Studio 上启用？

采纳答案by Necrolis

MSDNdoesn't list it as being available, and this recent responseagrees, so officially, no, there is no type called __int128and it can't be enabled.

MSDN没有将它列为可用，并且最近的这个回应同意，所以正式地，不，没有调用的类型__int128并且它不能被启用。

Additionally, never trust the syntax hilighter; it is user editable, and thus likely to either have bogus or 'future' types in it. (it is probably a reserved word however, due to the error, thus you should avoid naming any types __int128, this follows the convention that anything prefixed with a double underscore should reserved for compiler use).

此外，永远不要相信语法高亮器；它是用户可编辑的，因此可能包含虚假或“未来”类型。（但是，由于错误，它可能是一个保留字，因此您应该避免命名任何类型__int128，这遵循以下约定，即任何以双下划线为前缀的内容都应保留供编译器使用）。

One would think the __int128might be available on x64/IPF machines via register spanning, like __in64is on 32bit targets, but right now the only128 bit types stem from SIMD types (__m128and its various typed forms).

有人会认为它__int128可能通过寄存器跨越在 x64/IPF 机器上可用，就像__in64在 32 位目标上一样，但现在唯一的128 位类型源于 SIMD 类型（__m128及其各种类型的形式）。

回答by Kermit the Frog

There is a new version of _int128 which solves some of the problems mentioned. It includes a natvis addin, so you can view int128 in the debugger. To do this it was nessecary to write an x86 version of int128, because natvis-dll's need to be win32. The idea of using af template for the members lo,hi is ok, but I think it's a little to optimistic, because the routines that do the real job have to use the CPU's registers which, at least at the moment, are only 64 bit. But ok when Intel releases a 128-bit CPU. in/out in c++ std stream are added A lot of inline operators have been added too, so the compiler will do

_int128 的新版本解决了提到的一些问题。它包含一个 natvis 插件，因此您可以在调试器中查看 int128。为此，必须编写 x86 版本的 int128，因为 natvis-dll 需要是 win32。为成员 lo,hi 使用 af 模板的想法是可以的，但我认为这有点乐观，因为真正起作用的例程必须使用 CPU 的寄存器，至少目前只有 64 位. 但是当英特尔发布 128 位 CPU 时就可以了。c++ std 流中的 in/out 添加了很多内联运算符也添加了，所以编译器会做

_int128 x = 10;
int y = 20;
_int128 z = x + y;

without ambiguities.

没有歧义。

The code is too big to fit in this answer so it's put in github with links to the files listet below

代码太大，无法放入此答案中，因此将其放在 github 中，并附有指向以下文件列表的链接

New header Int128.h

新头文件 Int128.h

Int128x64.asm Assembler code for x64

Int128x64.asm x64 的汇编代码

Int128x86.cpp

Int128x86.cpp

Int128Str.cpp Common for x86 and x64

Int128Str.cpp x86 和 x64 通用

Int128IO.cpp Common for x86 and x64

Int128IO.cpp x86 和 x64 通用

AddIn-dll called by debugger to convert _int128/_uint128 to char*(radix 10)

调试器调用的 AddIn-dll 将 _int128/_uint128 转换为 char*(基数 10)

Header for all natvis addin dll's

所有 natvis 插件 dll 的标头

回答by Kermit the Frog

I've found a treasure in my old Visual Studio 6.0 C++ from 1996 (32-bit) making use of MS's own assembler routines that enabled 64-bit math on a 32-bit CPU(__int64). Source-code unfortunately not available). However, doing a debug-session that calls these functions, copy/paste the disassembler-list, search-replace "dword ptr" -> "qword ptr", eax,ebx,... -> rax,rbx,... and a Little adjustment of registers used for parameter-passing (and a lot of coffee), I succeeded to make this code, that makes it possible to do _int128-math in x64-mode just as it is possible to do __int64-math with 32-bit. It's essential the same code, with a doubleup in bits/cycle. For the matter of copyrights, I've seen no licenses in the disassembler-list, and perhaps it's time for Microsoft to integrate this into their x64 C++ compiler (vers. 2015) The code goes here

我在 1996 年（32 位）的旧 Visual Studio 6.0 C++ 中发现了一个宝藏，利用 MS 自己的汇编程序在 32 位 CPU（__int64）上启用了 64 位数学。不幸的是，源代码不可用）。但是，执行调用这些函数的调试会话，复制/粘贴反汇编程序列表，搜索替换“dword ptr” -> “qword ptr”，eax,ebx,... -> rax,rbx,...和用于参数传递的寄存器的一些调整（和很多咖啡），我成功地制作了这段代码，这使得在 x64 模式下执行 _int128-math 成为可能，就像可以用 __int64-math 32 位。重要的是相同的代码，位/周期加倍。关于版权问题，我在反汇编列表中没有看到许可证，也许是

// File:Int128.h
#pragma once

#include "PragmaLib.h" // contains #pragma comment(lib,"Yourlib.lib")

#ifndef _M_X64

#error Int128 is available only in x64 arhcitecture

#else

class _int128;
class _uint128;
extern "C" {    
  void int128sum( void *dst, const void *x, const void *y);
  void int128dif( void *dst, const void *x, const void *y);
  void int128mul( void *dst, const void *x, const void *y);
  void int128div( void *dst, const void *x, const void *y);
  void int128rem( void *dst, const void *x, const void *y);
  void int128neg( void *dst, const void *x);
  int  int128cmp(const void *n1, const void *n2);
  void uint128div( void *dst, const void *x, const void *y);
  void uint128rem( void *dst, const void *x, const void *y);
  int  uint128cmp(const void *n1, const void *n2);
};

class _int128 {
private:
  _int128(unsigned __int64 _lo, const unsigned __int64 _hi) : lo(_lo), hi(_hi) {
  }
public:
  unsigned __int64 lo;
  unsigned __int64 hi;

  inline _int128() {
  }
  inline _int128(unsigned __int64 n) : lo(n), hi(0) {
  }
  inline _int128(__int64 n) : lo(n), hi(n>=0)?0:-1) { // remember signextend hi if n < 0 (2-complement)
  }
  inline _int128(unsigned int n) : lo(n), hi(0) {
  }
  inline _int128(int n) : lo(n), hi(n>=0)?0:-1) {
  }
  inline _int128(unsigned short n) : lo(n), hi(0) {
  }
  inline _int128(short n) : lo(n), hi(n>=0)?0:-1) {
  }
  explicit _int128(const char *str);

  operator unsigned __int64() const {
    return lo;
  }
  operator __int64() const {
    return lo;
  }
  operator unsigned int() const {
    return (unsigned int)lo;
  }
  operator int() const {
    return (int)lo;
  }
  inline _int128 operator+(const _int128 &rhs) const {
    _int128 result;
    int128sum(&result, this, &rhs);
    return result;
  }

  inline _int128 operator-(const _int128 &rhs) const {
    _int128 result;
    int128dif(&result, this, &rhs);
    return result;
  }

  inline _int128 operator-() const {
    _int128 result;
    int128neg(&result, this);
    return result;
  }
  inline _int128 operator*(const _int128 &rhs) const {
    _int128 result;
    int128mul(&result, this, &rhs);
    return result;
  }

  inline _int128 operator/(const _int128 &rhs) const {
    _int128 result, copy(*this);
    int128div(&result, &copy, &rhs);
    return result;
  }
  inline _int128 operator%(const _int128 &rhs) const {
    _int128 result, copy(*this);
    int128rem(&result, &copy, &rhs);
    return result;
  };

  inline _int128 &operator+=(const _int128 &rhs) {
    const _int128 copy(*this);
    int128sum(this, &copy, &rhs);
    return *this;
  }
  inline _int128 &operator-=(const _int128 &rhs) {
    const _int128 copy(*this);
    int128dif(this, &copy, &rhs);
    return *this;
  }
  inline _int128 &operator*=(const _int128 &rhs) {
    const _int128 copy(*this);
    int128mul(this, &copy, &rhs);
    return *this;
  }
  inline _int128 &operator/=(const _int128 &rhs) {
    const _int128 copy(*this);
    int128div(this, &copy, &rhs);
    return *this;
  }
  inline _int128 &operator%=(const _int128 &rhs) {
    const _int128 copy(*this);
    int128rem(this, &copy, &rhs);
    return *this;
  }

  inline _int128 operator&(const _int128 &rhs) const {
    return _int128(lo&rhs.lo, hi&rhs.hi);
  }
  inline _int128 operator|(const _int128 &rhs) const {
    return _int128(lo|rhs.lo, hi|rhs.hi);
  }
  inline _int128 operator^(const _int128 &rhs) const {
    return _int128(lo^rhs.lo, hi^rhs.hi);
  }

  const char *parseDec(const char *str); // return pointer to char following the number
  const char *parseHex(const char *str); // do
  const char *parseOct(const char *str); // do
};

class _uint128 {
public:
  unsigned __int64 lo;
  unsigned __int64 hi;

  inline _uint128() {
  }
  inline _uint128(const _int128 &n) : lo(n.lo), hi(n.hi) {
  }
  inline _uint128(unsigned __int64 n) : lo(n), hi(0) {
  }
  inline _uint128(__int64 n) : lo(n), hi(n>=0)?0:-1) {
  }
  inline _uint128(unsigned int n) : lo(n), hi(0) {
  }
  inline _uint128(int n) : lo(n), hi(n>=0)?0:-1) {
  }
  inline _uint128(unsigned short n) : lo(n), hi(0) {
  }
  inline _uint128(short n) : lo(n), hi(n>=0)?0:-1) {
  }
  explicit _uint128(const char *str);

  inline operator _int128() const {
    return *(_int128*)(void*)this;
  }
  inline operator unsigned __int64() const {
    return lo;
  }
  inline operator __int64() const {
    return lo;
  }
  inline operator unsigned int() const {
    return (unsigned int)lo;
  }
  inline operator int() const {
    return (int)lo;
  }

  inline _uint128 operator+(const _uint128 &rhs) const {
    _uint128 result;
    int128sum(&result, this, &rhs);
    return result;
  }

  inline _uint128 operator-(const _uint128 &rhs) const {
    _uint128 result;
    int128dif(&result, this, &rhs);
    return result;
  }

  inline _uint128 operator*(const _uint128 &rhs) const {
    _uint128 result;
    int128mul(&result, this, &rhs);
    return result;
  }

  inline _uint128 operator/(const _uint128 &rhs) const {
    _uint128 result, copy(*this);
    uint128div(&result, &copy, &rhs);
    return result;
  }

  inline _uint128 operator%(const _uint128 &rhs) const {
    _uint128 result, copy(*this);
    uint128rem(&result, &copy, &rhs);
    return result;
  };

  inline _uint128 &operator+=(const _uint128 &rhs) {
    const _uint128 copy(*this);
    int128sum(this, &copy, &rhs);
    return *this;
  }
  inline _uint128 &operator-=(const _uint128 &rhs) {
    const _uint128 copy(*this);
    int128dif(this, &copy, &rhs);
    return *this;
  }
  inline _uint128 &operator*=(const _uint128 &rhs) {
    const _uint128 copy(*this);
    int128mul(this, &copy, &rhs);
    return *this;
  }
  inline _uint128 &operator/=(const _uint128 &rhs) {
    const _uint128 copy(*this);
    uint128div(this, &copy, &rhs);
    return *this;
  }
  inline _uint128 &operator%=(const _uint128 &rhs) {
    const _uint128 copy(*this);
    uint128rem(this, &copy, &rhs);
    return *this;
  }
  const char *parseDec(const char *str); // return pointer to char following the number
  const char *parseHex(const char *str); // do
  const char *parseOct(const char *str); // do

};

inline bool operator==(const _int128 &lft, const _int128 &rhs) {
  return (lft.lo == rhs.lo) && (lft.hi == rhs.hi);
}
inline bool operator==(const _int128 &lft, const _uint128 &rhs) {
  return (lft.lo == rhs.lo) && (lft.hi == rhs.hi);
}
inline bool operator==(const _uint128 &lft, const _int128 &rhs) {
  return (lft.lo == rhs.lo) && (lft.hi == rhs.hi);
}
inline bool operator==(const _uint128 &lft, const _uint128 &rhs) {
  return (lft.lo == rhs.lo) && (lft.hi == rhs.hi);
}
inline bool operator!=(const _int128 &lft, const _int128 &rhs) {
  return (lft.lo != rhs.lo) || (lft.hi != rhs.hi);
}
inline bool operator!=(const _int128 &lft, const _uint128 &rhs) {
  return (lft.lo != rhs.lo) || (lft.hi != rhs.hi);
}
inline bool operator!=(const _uint128 &lft, const _int128 &rhs) {
  return (lft.lo != rhs.lo) || (lft.hi != rhs.hi);
}
inline bool operator!=(const _uint128 &lft, const _uint128 &rhs) {
  return (lft.lo != rhs.lo) || (lft.hi != rhs.hi);
}

inline bool operator>(const _int128 &lft, const _int128 &rhs) {
  return int128cmp(&lft, &rhs) > 0;
}
inline bool operator>(const _int128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) > 0;
}
inline bool operator>(const _uint128 &lft, const _int128 &rhs) {
  return uint128cmp(&lft, &rhs) > 0;
}
inline bool operator>(const _uint128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) > 0;
}

inline bool operator>=(const _int128 &lft, const _int128 &rhs) {
  return int128cmp(&lft, &rhs) >= 0;
}
inline bool operator>=(const _int128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) >= 0;
}
inline bool operator>=(const _uint128 &lft, const _int128 &rhs) {
  return uint128cmp(&lft, &rhs) >= 0;
}
inline bool operator>=(const _uint128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) >= 0;
}

inline bool operator<(const _int128 &lft, const _int128 &rhs) {
  return int128cmp(&lft, &rhs) < 0;
}
inline bool operator<(const _int128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) < 0;
}
inline bool operator<(const _uint128 &lft, const _int128 &rhs) {
  return uint128cmp(&lft, &rhs) < 0;
}
inline bool operator<(const _uint128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) < 0;
}

inline bool operator<=(const _int128 &lft, const _int128 &rhs) {
  return int128cmp(&lft, &rhs) <= 0;
}
inline bool operator<=(const _int128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) <= 0;
}
inline bool operator<=(const _uint128 &lft, const _int128 &rhs) {
  return uint128cmp(&lft, &rhs) <= 0;
}
inline bool operator<=(const _uint128 &lft, const _uint128 &rhs) {
  return uint128cmp(&lft, &rhs) <= 0;
}

char    * _i128toa(_int128 value, char *str, int radix);
char    * _ui128toa(_uint128 value, char *str, int radix);
wchar_t * _i128tow(_int128 value, wchar_t *str, int radix);
wchar_t * _ui128tow(_uint128 value, wchar_t *str, int radix);

inline char radixLetter(unsigned int c) {
  return (c < 10) ? ('0' + c) : ('a' + (c-10));
}

inline wchar_t wradixLetter(unsigned int c) {
  return (c < 10) ? ('0' + c) : ('a' + (c-10));
}

inline bool isodigit(unsigned char ch) {
  return ('0' <= ch) && (ch < '8');
}

unsigned int convertNumberChar(char digit);

#endif // _M_X64

; File: Int128x64.asm
; build obj-file with
; ml64 /nologo /c /Zf /Fo$(IntDir)Int128x64.obj Int128x64.asm
.CODE

;void int128sum(_int128 &dst, cnost _int128 &x, const _int128 &y);
int128sum PROC
    push    rbx
    mov     rax, qword ptr[rdx]
    add     rax, qword ptr[r8]
    mov     rbx, qword ptr[rdx+8]
    adc     rbx, qword ptr[r8+8]
    mov     qword ptr[rcx], rax
    mov     qword ptr[rcx+8], rbx
    pop     rbx
    ret
int128sum ENDP

;void int128dif( _int128 &dst, const _int128 &x, const _int128 &y);
int128dif PROC
    push    rbx
    mov     rax, qword ptr[rdx]
    sub     rax, qword ptr[r8]
    mov     rbx, qword ptr[rdx+8]
    sbb     rbx, qword ptr[r8+8]
    mov     qword ptr[rcx]  , rax
    mov     qword ptr[rcx+8], rbx
    pop     rbx
    ret
int128dif ENDP

;void int128mul(_int128 &dst, const _int128 &x, const _int128 &y);
int128mul PROC
    push    rbx
    mov     rax, qword ptr[rdx+8]           ; rax = x.hi
    mov     rbx, qword ptr[r8+8]            ; rbx = y.hi
    or      rbx, rax                        ; rbx = x.hi | y.hi
    mov     rbx, qword ptr[r8]              ; rbx = y.lo
    jne     Hard                            ; if(x.hi|y.hi) goto Hard
                                            ; simple int64 multiplication
    mov     rax, qword ptr[rdx]             ; rax = x.lo
    mul     rbx                             ; rdx:rax = rax * rbx
    mov     qword ptr[rcx]  , rax           ; dst.lo = rax
    mov     qword ptr[rcx+8], rdx           ; dst.hi = rdx
    pop     rbx
    ret
Hard:                                       ; assume rax = x.hi, rbx = y.lo
    push    rsi
    mov     rsi, rdx                        ; need rdx for highend of mul, so rsi=&x
    mul     rbx                             ; rdx:rax = x.hi * y.lo
    mov     r9 , rax                        ; 
    mov     rax, qword ptr[rsi]             ; rax     = x.lo
    mul     qword ptr[r8+8]                 ; rdx:rax = x.lo * y.hi
    add     r9, rax                         ; r9      = lo(x.hi*y.lo+x.lo*y.hi); 
    mov     rax, qword ptr[rsi]             ; rax     = x.lo
    mul     rbx                             ; rdx:rax = x.lo * y.lo
    add     rdx, r9
    mov     qword ptr[rcx]  , rax
    mov     qword ptr[rcx+8], rdx
    pop     rsi
    pop     rbx
    ret
int128mul ENDP


;void int128div(_int128 &dst, const _int128 &x, const _int128 &y);
int128div PROC
    push        rdi
    push        rsi
    push        rbx
    push        rcx
    mov         r9,  rdx
    xor         rdi, rdi
    mov         rax, qword ptr[r9+8]
    or          rax, rax
    jge         L1
    inc         rdi
    mov         rdx, qword ptr[r9]
    neg         rax
    neg         rdx
    sbb         rax, 0
    mov         qword ptr[r9+8], rax
    mov         qword ptr[r9], rdx
L1:
    mov         rax, qword ptr[r8+8]
    or          rax, rax
    jge         L2
    inc         rdi
    mov         rdx, qword ptr[r8]
    neg         rax
    neg         rdx
    sbb         rax,0
    mov         qword ptr[r8+8], rax
    mov         qword ptr[r8], rdx
L2:
    or          rax, rax
    jne         L3
    mov         rcx, qword ptr[r8]
    mov         rax, qword ptr[r9+8]
    xor         rdx, rdx
    div         rcx
    mov         rbx, rax
    mov         rax, qword ptr[r9]
    div         rcx
    mov         rdx, rbx
    jmp         L4
L3:
    mov         rbx,rax
    mov         rcx,qword ptr[r8]
    mov         rdx,qword ptr[r9+8]
    mov         rax,qword ptr[r9]
L5:
    shr         rbx, 1
    rcr         rcx, 1
    shr         rdx, 1
    rcr         rax, 1
    or          rbx, rbx
    jne         L5
    div         rcx
    mov         rsi, rax
    mul         qword ptr[r8+8]
    mov         rcx, rax
    mov         rax, qword ptr[r8]
    mul         rsi
    add         rdx, rcx
    jb          L6
    cmp         rdx, qword ptr[r9+8]
    ja          L6
    jb          L7
    cmp         rax, qword ptr[rdx]
    jbe         L7
L6:
    dec         rsi
L7:
    xor         rdx, rdx
    mov         rax, rsi
L4:
    dec         rdi
    jne         L8
    neg         rdx
    neg         rax
    sbb         rdx, 0
L8:
    pop         rcx
    pop         rbx
    pop         rsi
    pop         rdi
    mov         qword ptr[rcx], rax
    mov         qword ptr[rcx+8], rdx
    ret
int128div ENDP

;void int128rem( _int128 &dst, const _int128 &x, const _int128 &y);
int128rem PROC
    push        rbx
    push        rdi
    push        rcx
    mov         r9,  rdx
    xor         rdi, rdi
    mov         rax, qword ptr[r9+8]
    or          rax, rax
    jge         L1
    inc         rdi
    mov         rdx, qword ptr[r9]
    neg         rax
    neg         rdx
    sbb         rax, 0
    mov         qword ptr[r9+8], rax
    mov         qword ptr[r9], rdx
L1:
    mov         rax, qword ptr[r8+8]
    or          rax, rax
    jge         L2
    mov         rdx, qword ptr[r8]
    neg         rax
    neg         rdx
    sbb         rax, 0
    mov         qword ptr[r8+8], rax
    mov         qword ptr[r8], rdx
L2:
    or          rax, rax
    jne         L3
    mov         rcx, qword ptr[r8]
    mov         rax, qword ptr[r9+8]
    xor         rdx, rdx
    div         rcx
    mov         rax, qword ptr[r9]
    div         rcx
    mov         rax, rdx
    xor         rdx, rdx
    dec         rdi
    jns         L4
    jmp         L8
L3:
    mov         rbx, rax
    mov         rcx, qword ptr[r8]
    mov         rdx, qword ptr[r9+8]
    mov         rax, qword ptr[r9]
L5:
    shr         rbx, 1
    rcr         rcx, 1
    shr         rdx, 1
    rcr         rax, 1
    or          rbx, rbx
    jne         L5
    div         rcx
    mov         rcx, rax
    mul         qword ptr[r8+8]
    xchg        rax, rcx
    mul         qword ptr[r8]
    add         rdx, rcx
    jb          L6
    cmp         rdx, qword ptr[r9+8]
    ja          L6
    jb          L7
    cmp         rax, qword ptr[r9]
    jbe         L7
L6:
    sub         rax, qword ptr[r8]
    sbb         rdx, qword ptr[r8+8]
L7:
    sub         rax, qword ptr[r9]
    sbb         rdx, qword ptr[r9+8]
    dec         rdi
    jns         L8
L4:
    neg         rdx
    neg         rax
    sbb         rdx, 0
L8:
    pop         rcx
    pop         rdi
    pop         rbx
    mov         qword ptr[rcx], rax
    mov         qword ptr[rcx+8], rdx
    ret
int128rem ENDP

;void int128neg( _int128 &dst, const _int128 &x);
int128neg PROC
    mov         rax,qword ptr[rdx]
    neg         rax
    mov         r8, qword ptr[rdx+8]
    adc         r8, 0
    neg         r8
    mov         qword ptr[rcx], rax
    mov         qword ptr[rcx+8], r8
    ret
int128neg ENDP

;int int128cmp(const _int128 &n1, const _int128 &n2);
int128cmp PROC
    mov         rax, qword ptr[rcx+8]       ; n1.hi
    cmp         rax, qword ptr[rdx+8]       ; n2.hi
    jl          lessthan                            ; signed compare of n1.hi and n2.hi
    jg          greaterthan
    mov         rax, qword ptr[rcx]         ; n2.lo
    cmp         rax, qword ptr[rdx]         ; n2.lo
    jb          lessthan                    ; unsigned compare of n1.lo and n2.lo
    ja          greaterthan
    mov         rax, 0                      ; they are equal
    ret
greaterthan:
    mov         rax, 1
    ret
lessthan:
    mov         rax, -1
    ret
int128cmp ENDP

END

; File:UInt128x64.asm
; build obj-file with
; ml64 /nologo /c /Zf /Fo$(IntDir)UInt128x64.obj UInt128x64.asm

.CODE

;void uint128div(_uint128 &dst, const _uint128 &x, const _uint128 &y);
uint128div PROC
    push        rbx
    push        rsi
    push        rcx
    mov         r9, rdx
    mov         rax, qword ptr[r8+8]
    or          rax, rax
    jne         L1
    mov         rcx, qword ptr[r8]
    mov         rax, qword ptr[r9+8]
    xor         rdx, rdx
    div         rcx
    mov         rbx, rax
    mov         rax, qword ptr[r9]
    div         rcx
    mov         rdx, rbx
    jmp         L2
L1:
    mov         rcx, rax
    mov         rbx, qword ptr[r8]
    mov         rdx, qword ptr[r9+8]
    mov         rax, qword ptr[r9]
L3:
    shr         rcx, 1
    rcr         rbx, 1
    shr         rdx, 1
    rcr         rax, 1
    or          rcx, rcx
    jne         L3
    div         rbx
    mov         rsi, rax
    mul         qword ptr[r8+8]
    mov         rcx, rax
    mov         rax, qword ptr[r8]
    mul         rsi
    add         rdx, rcx
    jb          L4
    cmp         rdx, qword ptr[r9+8]
    ja          L4
    jb          L5
    cmp         rax, qword ptr[r9]
    jbe         L5
L4:
    dec         rsi
L5:
    xor         rdx, rdx
    mov         rax, rsi
L2:
    pop         rcx
    pop         rsi
    pop         rbx
    mov         qword ptr[rcx], rax
    mov         qword ptr[rcx+8], rdx
    ret
uint128div ENDP

;void uint128rem(_uint128 &dst, const _uint128 &x, const _uint128 &y);
uint128rem PROC
    push        rbx
    push        rcx
    mov         r9, rdx
    mov         rax, qword ptr[r8+8]
    or          rax, rax
    jne         L1
    mov         rcx, qword ptr[r8]
    mov         rax, qword ptr[r9+8]
    xor         rdx, rdx
    div         rcx
    mov         rax, qword ptr[r9]
    div         rcx
    mov         rax, rdx
    xor         rdx, rdx
    jmp         L2
L1:
    mov         rcx, rax
    mov         rbx, qword ptr[r8]
    mov         rdx, qword ptr[r9+8]
    mov         rax, qword ptr[r9]
L3:
    shr         rcx, 1
    rcr         rbx, 1
    shr         rdx, 1
    rcr         rax, 1
    or          rcx, rcx
    jne         L3
    div         rbx
    mov         rcx, rax
    mul         qword ptr[r8+8]
    xchg        rax, rcx
    mul         qword ptr[r8]
    add         rdx, rcx
    jb          L4
    cmp         rdx, qword ptr[r9+8]
    ja          L4
    jb          L5
    cmp         rax, qword ptr[r9]
    jbe         L5
L4:
    sub         rax, qword ptr[r8]
    sbb         rdx, qword ptr[r8+8]
L5:
    sub         rax, qword ptr[r9]
    sbb         rdx, qword ptr[r9+8]
    neg         rdx
    neg         rax
    sbb         rdx, 0
L2:
    pop         rcx
    pop         rbx
    mov         qword ptr[rcx], rax
    mov         qword ptr[rcx+8], rdx
    ret
uint128rem ENDP

;int uint128cmp(const _uint128 &n1, const _uint128 &n2);
uint128cmp PROC
    mov         rax, qword ptr[rcx+8]       ; n1.hi
    cmp         rax, qword ptr[rdx+8]       ; n2.hi
    jb          lessthan                    ; usigned compare of n1.hi and n2.hi
    ja          greaterthan
    mov         rax, qword ptr[rcx]         ; n2.lo
    cmp         rax, qword ptr[rdx]         ; n2.lo
    jb          lessthan                    ; unsigned compare of n1.lo and n2.lo
    ja          greaterthan
    mov         rax, 0                      ; they are equal
    ret
greaterthan:
    mov         rax, 1
    ret
lessthan:
    mov         rax, -1
    ret
uint128cmp ENDP

END

There will be 3 more files. not enough Space here...

还会有 3 个文件。这里没有足够的空间...

回答by Kermit the Frog

And the rest is here. (conversion functions to/from string)

剩下的就在这里。（转换函数到/从字符串）

// File:Int128IOx64.cpp
#include "pch.h"

#ifdef _M_X64

#include <Math/Int128.h>

static const _int128 _0(0);
static const _int128 _10(10);
static const _int128 _16(16);
static const _int128 _8(16);

char *_i128toa(_int128 value, char *str, int radix) {
  assert(radix >= 2 && radix <= 36);
  char *s = str;
  const bool negative = value < _0;
  if (negative && (radix == 10)) {
    value = -value;
    while (value != _0) {
      const unsigned int c = value % _10;
      *(s++) = radixLetter(c);
      value /= _10;
    }
    *(s++) = '-';
    *s = 0;
    return _strrev(str);
  }

  _uint128 v(value);
  const _uint128 r(radix);
  while (v != _0) {
    const unsigned int c = v % r;
    *(s++) = radixLetter(c);
    v /= r;
  }
  if (s == str) {
    return strcpy(str, "0");
  }
  else {
    *s = 0;
    return _strrev(str);
  }
  return str;
}

wchar_t *_i128tow(_int128 value, wchar_t *str, int radix) {
  wchar_t *s = str;
  const bool negative = value < _0;
  if (negative && (radix == 10)) {
    value = -value;
    while (value != _0) {
      const unsigned int c = value % _10;
      *(s++) = wradixLetter(c);
      value /= _10;
    }
    *(s++) = '-';
    *s = 0;
    return _wcsrev(str);
  }

  _uint128 v(value);
  const _uint128 r(radix);
  while (v != _0) {
    const unsigned int c = v % r;
    *(s++) = radixLetter(c);
    v /= r;
  }
  if (s == str) {
    return wcscpy(str, L"0");
  }
  else {
    *s = 0;
    return _wcsrev(str);
  }
  return str;
}

const char *_int128::parseDec(const char *str) { // return pointer to char following the number
  bool negative = false;
  bool gotDigit = false;
  switch (*str) {
  case '+':
    str++;
    break;
  case '-':
    str++;
    negative = true;
  }
  *this = _0;
  while (isdigit(*str)) {
    gotDigit = true;
    const unsigned int d = *(str++) - '0';
    *this *= _10;
    *this += d;
  }
  if (!gotDigit) {
    throw "_int128:string is not a number";
  }
  if (negative) {
    *this = -*this;
  }
  return str;
}

const char *_int128::parseHex(const char *str) {
  *this = 0;
  while (isxdigit(*str)) {
    const unsigned int d = convertNumberChar(*(str++));
    *this *= _16;
    *this += d;
  }
  return str;
}

const char *_int128::parseOct(const char *str) {
  *this = 0;
  while (isodigit(*str)) {
    const unsigned int d = convertNumberChar(*(str++));
    *this *= _8;
    *this += d;
  }
  return str;
}

_int128::_int128(const char *str) {
  if (*str == '-') {
    parseDec(str);
  } else {
    if (!isdigit(*str)) {
      throw exception("_int128:string is not an integer");
    }
    if (*str == '0') {
      switch (str[1]) {
      case 'x':
        parseHex(str + 2);
        break;
      case 0:
        *this = 0;
        break;
      default:
        parseOct(str + 1);
      }
    }
    else {
      parseDec(str);
    }
  }
}
#endif // _M_X64


// File:UInt128IOx64.cpp
#include "pch.h"

#ifdef _M_X64

#include <Math/Int128.h>

static const _uint128 _0(0);
static const _uint128 _10(10);
static const _uint128 _16(16);
static const _uint128 _8(16);

char*_ui128toa(_uint128 value, char *str, int radix) {
  assert(radix >= 2 && radix <= 36);
  char *s = str;
  const _uint128 r(radix);
  while (value != _0) {
    const unsigned int c = value % r;
    *(s++) = radixLetter(c);
    value /= r;
  }
  if (s == str) {
    return strcpy(str, "0");
  }
  else {
    *s = 0;
    return _strrev(str);
  }
}

wchar_t *_ui128tow(_uint128 value, wchar_t *str, int radix) {
  assert(radix >= 2 && radix <= 36);
  wchar_t *s = str;
  const _uint128 r(radix);
  while (value != _0) {
    const unsigned int c = value % r;
    *(s++) = wradixLetter(c);
    value /= r;
  }
  if (s == str) {
    return wcscpy(str, L"0");
  }
  else {
    *s = 0;
    return _wcsrev(str);
  }
}

const char *_uint128::parseDec(const char *str) {
  *this = 0;
  while (isdigit(*str)) {
    const unsigned int d = *(str++) - '0';
    *this *= _10;
    *this += d;
  }
  return str;
}

const char *_uint128::parseHex(const char *str) {
  *this = 0;
  while (isxdigit(*str)) {
    const unsigned int d = convertNumberChar(*(str++));
    *this *= _16;
    *this += d;
  }
  return str;
}

const char *_uint128::parseOct(const char *str) {
  *this = 0;
  while (isodigit(*str)) {
    const unsigned int d = convertNumberChar(*(str++));
    *this *= _8;
    *this += d;
  }
  return str;
}

_uint128::_uint128(const char *str) {
  if (!isdigit(*str)) {
    throw exception("_uint128:string is not an integer");
  }
  if (*str == '0') {
    switch (str[1]) {
    case 'x':
      parseHex(str + 2);
      break;
    case 0:
      *this = 0;
      break;
    default:
      parseOct(str + 1);
      break;
    }
  }
  else {
    parseDec(str);
  }
}

#endif // _M_X64

// File:Int128IOCommon.cpp
#include "pch.h"

#ifdef _M_X64

#include <Math/Int128.h>

unsigned int convertNumberChar(char digit) {
  switch(digit) {
  case '0': return 0;
  case '1': return 1;
  case '2': return 2;
  case '3': return 3;
  case '4': return 4;
  case '5': return 5;
  case '6': return 6;
  case '7': return 7;
  case '8': return 8;
  case '9': return 9;
  case 'a':
  case 'A': return 10;
  case 'b':
  case 'B': return 11;
  case 'c':
  case 'C': return 12;
  case 'd':
  case 'D': return 13;
  case 'e':
  case 'E': return 14;
  case 'f':
  case 'F': return 15;
  default :
    return 0;
  }
}

#endif // _M_X64

回答by BenGoldberg

Your conversions to/from string could use some improvement.

您与字符串之间的转换可能需要一些改进。

For saftey, the interface for converting to a string should include the allocated length of the user provided string, so you can return an error if they didn't provide enough memory.

为了安全起见，转换为字符串的接口应该包含用户提供的字符串的分配长度，因此如果它们没有提供足够的内存，您可以返回错误。

Also, try to process strings in chunks: for example, suppose the user wants to convert a 128 bit number into base 10. Instead of repeatedly doing modulo 10, you could be doing modulo 1000000000ul, and using sprintf(s, "%09u", c).

此外，尝试分块处理字符串：例如，假设用户想要将 128 位数字转换为以 10 为基数的数字。而不是重复执行模 10，您可以执行 modulo 1000000000ul，并使用sprintf(s, "%09u", c).

Converting from a string could be similarly optimized.

从字符串转换可以类似地优化。

It would not be a bad idea to include a divremmethod, with a return type of std::pair<_uint128, _uint128>.

包含divrem一个返回类型为std::pair<_uint128, _uint128>.

It would be absolutely awesome if you had a integer class where the type used for hiand lowas a template parameter. Then, with a small handful of typedefs, you could create an int256, an int512, etc..

如果您有一个整数类，其中的类型用于hi并且lo是模板参数，那将是非常棒的。然后，使用少量 typedef，您可以创建一个 int256、一个 int512 等。