Assigning 128-bit integer in C

Question

When I try to assign an 128-bit integer in gcc 4.9.1, I get a warning: integer constant is too large for its type.

Example Code

int main(void) {
  __uint128_t p = 47942806932686753431;

  return 0;
}

Output

I'm compiling with gcc -std=c11 -o test test.c and I get:

test.c: In function ‘main’:
test.c:2:19: warning: integer constant is too large for its type
   __uint128_t p = 47942806932686753431;
               ^

Am I doing something wrong or is this a bug in gcc?

Answer 1

Am I doing something wrong or is this a bug in gcc?

The problem is in 47942806932686753431 part, not in __uint128_t p. According to gcc docs there's no way to declare 128 bit constant:

There is no support in GCC for expressing an integer constant of type __int128 for targets with long long integer less than 128 bits wide.

So, it seems that while you can have 128 bit variables, you cannot have 128 bit constants, unless your long long is 128 bit wide.

The workaround could be to construct 128 bit value from "narrower" integral constants using basic arithmetic operations, and hope for compiler to perform constant folding.

Answer 2

Have you tried this?

__int128 p = *(__int128*) "\x00\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0a\x0b\x0c\x0d\x0e\x0f";

EDIT Nov. 25

Sorry for the poor clarification on previous post. Seriously, I didn't post this answer as a joke. Though the GCC doc states there's no way to express a 128-bit integer constant, this post simply provides a workaround for those who wants to assign values to __uint128_t variables with ease.

You may try to compile the code below with GCC (7.2.0) or Clang (5.0.0). It prints desired results.

#include <stdint.h>
#include <stdio.h>

int main()
{
    __uint128_t p = *(__int128*) "\x00\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0a\x0b\x0c\x0d\x0e\x0f";
    printf("HIGH %016llx\n", (uint64_t) (p >> 64));
    printf("LOW  %016llx\n", (uint64_t) p);
    return 0;
}

The stdout:

HIGH 0f0e0d0c0b0a0908
LOW  0706050403020100

This is only regarded as a workaround since it plays tricks on pointers by placing the "value" in .rodata section (if you objdump it), and it's not portable (x86_64 and aarch64 are fine but not arm and x86). I think it's been enough for those coding on desktop machines.

Answer 3

I'd suggest a simple macro to combine two 64-bit values:

#define UINT128(hi, lo) (((__uint128_t) (hi)) << 64 | (lo))

Example usage:

#include <stdint.h>
#include <stdio.h>

#define UINT128(hi, lo) (((__uint128_t) (hi)) << 64 | (lo))

int main() {
  // Creates constant 0x010203040506070890a0b0c0d0e0f10
  __uint128_t x = UINT128(0x0102030405060708, 0x090a0b0c0d0e0f10);

  // Prints 1, 2, 3, 4, .. 15, 16
  for (int i = 0; i < 16; ++i) {
    printf("%d ", (int)((x >> (120 - 8*i)) & 255));
  }
  printf("\n");
}

Alternatively, you could move the 0x prefix into the macro (not recommended):

#define UINT128(hi, lo) (((__uint128_t) (0x##hi)) << 64 | (0x##lo))

So you can do UINT128(0102030405060708, 090a0b0c0d0e0f10) which is slightly shorter, but that might mess up syntax highlighting, because 0102030405060708 could be misinterpreted as an octal constant, and 090a0b0c0d0e0f10 is not a valid token by itself, which is why I wouldn't recommend this.

Answer 4

I had same issue and cooked up a solution using user-defined literals. Here is how you instantiate the user-defined literal _xxl:

int main(int argc, char** argv) {

    auto a = 0xF0000000000000000000000000000000LLU;
    auto b = 0xF0000000000000000000000000000000_xxl;
    
    printf("sizeof(a): %zu\n", sizeof(a));
    printf("sizeof(b): %zu\n", sizeof(b));

    printf("a == 0? %s\n", a==0 ? "true":"false");
    printf("b == 0? %s\n", b==0 ? "true":"false");

    printf("b >> 124 = %x\n", b >> 124);

    return 0;
}

Output:

sizeof(a): 8
sizeof(b): 16
a == 0? true
b == 0? false
b >> 124 = f

Here is the implementation for the user-defined literal _xxl

#pragma once

#include <stdint.h>

#ifdef __SIZEOF_INT128__
    using uint_xxl_t = __uint128_t;
    using sint_xxl_t = __int128_t;
    
namespace detail_xxl
{
    constexpr uint8_t hexval(char c) 
    { return c>='a' ? (10+c-'a') : c>='A' ? (10+c-'A') : c-'0'; }

    template <int BASE, uint_xxl_t V>
    constexpr uint_xxl_t lit_eval() { return V; }
    
    template <int BASE, uint_xxl_t V, char C, char... Cs>
    constexpr uint_xxl_t lit_eval() {
        static_assert( BASE!=16 || sizeof...(Cs) <=  32-1, "Literal too large for BASE=16");
        static_assert( BASE!=10 || sizeof...(Cs) <=  39-1, "Literal too large for BASE=10");
        static_assert( BASE!=8  || sizeof...(Cs) <=  44-1, "Literal too large for BASE=8");
        static_assert( BASE!=2  || sizeof...(Cs) <= 128-1, "Literal too large for BASE=2");
        return lit_eval<BASE, BASE*V + hexval(C), Cs...>();
    }
    
    template<char... Cs > struct LitEval 
    {static constexpr uint_xxl_t eval() {return lit_eval<10,0,Cs...>();} };

    template<char... Cs> struct LitEval<'0','x',Cs...> 
    {static constexpr uint_xxl_t eval() {return lit_eval<16,0,Cs...>();} };

    template<char... Cs> struct LitEval<'0','b',Cs...> 
    {static constexpr uint_xxl_t eval() {return lit_eval<2,0,Cs...>();} };

    template<char... Cs> struct LitEval<'0',Cs...> 
    {static constexpr uint_xxl_t eval() {return lit_eval<8,0,Cs...>();} };
    
    template<char... Cs> 
    constexpr uint_xxl_t operator "" _xxl() {return LitEval<Cs...>::eval();}
}

template<char... Cs> 
constexpr uint_xxl_t operator "" _xxl() {return ::detail_xxl::operator "" _xxl<Cs...>();}
    
#endif // __SIZEOF_INT128__

It can be used in constexpr just like normal integer constants:

static_assert(   0_xxl == 0, "_xxl error" );
static_assert( 0b0_xxl == 0, "_xxl error" );
static_assert(  00_xxl == 0, "_xxl error" );
static_assert( 0x0_xxl == 0, "_xxl error" );

static_assert(   1_xxl == 1, "_xxl error" );
static_assert( 0b1_xxl == 1, "_xxl error" );
static_assert(  01_xxl == 1, "_xxl error" );
static_assert( 0x1_xxl == 1, "_xxl error" );

static_assert(      2_xxl == 2, "_xxl error" );
static_assert(   0b10_xxl == 2, "_xxl error" );
static_assert(     02_xxl == 2, "_xxl error" );
static_assert(    0x2_xxl == 2, "_xxl error" );

static_assert(      9_xxl == 9, "_xxl error" );
static_assert( 0b1001_xxl == 9, "_xxl error" );
static_assert(    011_xxl == 9, "_xxl error" );
static_assert(    0x9_xxl == 9, "_xxl error" );

static_assert(     10_xxl == 10, "_xxl error" );
static_assert(    0xa_xxl == 10, "_xxl error" );
static_assert(    0xA_xxl == 10, "_xxl error" );

static_assert( 0xABCDEF_xxl == 0xABCDEF, "_xxl error" );
static_assert( 1122334455667788_xxl == 1122334455667788LLu, "_xxl error" );
static_assert(0x80000000000000000000000000000000_xxl >> 126 == 0b10, "_xxl error");
static_assert(0x80000000000000000000000000000000_xxl >> 127 == 0b01, "_xxl error");
static_assert( 0xF000000000000000B000000000000000_xxl > 0xB000000000000000, "_xxl error" );