Add 64-bit emulation for platforms that don't have a native 64-bit type.

This commit is contained in:
Jordan Bancino 2023-08-12 16:30:24 +00:00
parent 098592de42
commit 92b1e60c4c
4 changed files with 634 additions and 7 deletions

264
src/Int64.c Normal file
View File

@ -0,0 +1,264 @@
/*
* Copyright (C) 2022-2023 Jordan Bancino <@jordan:bancino.net>
*
* Permission is hereby granted, free of charge, to any person
* obtaining a copy of this software and associated documentation files
* (the "Software"), to deal in the Software without restriction,
* including without limitation the rights to use, copy, modify, merge,
* publish, distribute, sublicense, and/or sell copies of the Software,
* and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include <Int64.h>
#include <stddef.h>
#include <signal.h>
size_t
UInt64Str(UInt64 x, int base, char *out, size_t len)
{
static const char symbols[] = "0123456789ABCDEF";
size_t i = len - 1;
size_t j = 0;
UInt64 base64 = UInt64Create(0, base);
/* We only have symbols up to base 16 */
if (base < 2 || base > 16)
{
return 0;
}
do
{
UInt64 mod = UInt64Rem(x, base64);
UInt32 low = UInt64Low(mod);
out[i] = symbols[low];
i--;
x = UInt64Div(x, base64);
} while (UInt64Gt(x, UInt64Create(0, 0)));
while (++i < len)
{
out[j++] = out[i];
}
out[j] = '\0';
return j;
}
#ifndef UINT64_NATIVE
/* No native 64-bit support, add our own */
UInt64
UInt64Create(UInt32 high, UInt32 low)
{
UInt64 x;
x.i[0] = low;
x.i[1] = high;
return x;
}
UInt64
UInt64Add(UInt64 x, UInt64 y)
{
UInt64 z = UInt64Create(0, 0);
int carry;
z.i[0] = x.i[0] + y.i[0];
carry = z.i[0] < x.i[0];
z.i[1] = x.i[1] + y.i[1] + carry;
return z;
}
UInt64
UInt64Sub(UInt64 x, UInt64 y)
{
UInt64 twosCompl = UInt64Add(UInt64Not(y), UInt64Create(0, 1));
return UInt64Add(x, twosCompl);
}
UInt64
UInt64Mul(UInt64 x, UInt64 y)
{
UInt64 z = UInt64Create(0, 0);
/* while (y > 0) */
while (UInt64Gt(y, UInt64Create(0, 0)))
{
/* if (y & 1 != 0) */
if (UInt64Neq(UInt64And(y, UInt64Create(0, 1)), UInt64Create(0, 0)))
{
z = UInt64Add(z, x);
}
x = UInt64Sll(x, 1);
y = UInt64Srl(y, 1);
}
return z;
}
typedef struct
{
UInt64 q;
UInt64 r;
} UInt64Ldiv;
static UInt64Ldiv
UInt64LongDivision(UInt64 n, UInt64 d)
{
UInt64Ldiv o;
int i;
o.q = UInt64Create(0, 0);
o.r = UInt64Create(0, 0);
if (UInt64Eq(d, UInt64Create(0, 0)))
{
raise(SIGFPE);
return o;
}
for (i = 63; i >= 0; i--)
{
UInt64 bit = UInt64And(UInt64Srl(n, i), UInt64Create(0, 1));
o.r = UInt64Sll(o.r, 1);
o.r = UInt64Or(o.r, bit);
if (UInt64Geq(o.r, d))
{
o.r = UInt64Sub(o.r, d);
o.q = UInt64Or(o.q, UInt64Sll(UInt64Create(0, 1), i));
}
}
return o;
}
UInt64
UInt64Div(UInt64 x, UInt64 y)
{
return UInt64LongDivision(x, y).q;
}
UInt64
UInt64Rem(UInt64 x, UInt64 y)
{
return UInt64LongDivision(x, y).r;
}
UInt64
UInt64Sll(UInt64 x, int y)
{
UInt64 z;
if (!y)
{
return x;
}
z = UInt64Create(0, 0);
if (y < 32)
{
z.i[1] = (x.i[0] >> (32 - y)) | (x.i[1] << y);
z.i[0] = x.i[0] << y;
}
else
{
z.i[1] = x.i[0] << (y - 32);
}
return z;
}
UInt64
UInt64Srl(UInt64 x, int y)
{
UInt64 z;
if (!y)
{
return x;
}
z = UInt64Create(0, 0);
if (y < 32)
{
z.i[0] = (x.i[1] << (32 - y)) | (x.i[0] >> y);
z.i[1] = x.i[1] >> y;
}
else
{
z.i[0] = x.i[1] >> (y - 32);
}
return z;
}
UInt64
UInt64And(UInt64 x, UInt64 y)
{
return UInt64Create(x.i[1] & y.i[1], x.i[0] & y.i[0]);
}
UInt64
UInt64Or(UInt64 x, UInt64 y)
{
return UInt64Create(x.i[1] | y.i[1], x.i[0] | y.i[0]);
}
UInt64
UInt64Xor(UInt64 x, UInt64 y)
{
return UInt64Create(x.i[1] ^ y.i[1], x.i[0] ^ y.i[0]);
}
UInt64
UInt64Not(UInt64 x)
{
return UInt64Create(~(x.i[1]), ~(x.i[0]));
}
int
UInt64Eq(UInt64 x, UInt64 y)
{
return x.i[0] == y.i[0] && x.i[1] == y.i[1];
}
int
UInt64Lt(UInt64 x, UInt64 y)
{
return x.i[1] < y.i[1] || (x.i[1] == y.i[1] && x.i[0] < y.i[0]);
}
int
UInt64Gt(UInt64 x, UInt64 y)
{
return x.i[1] > y.i[1] || (x.i[1] == y.i[1] && x.i[0] > y.i[0]);
}
#endif

View File

@ -55,10 +55,9 @@
#include <limits.h>
#define BIT64_MAX 18446744073709551615
#define BIT32_MAX 4294967295
#define BIT16_MAX 65535
#define BIT8_MAX 255
#define BIT32_MAX 4294967295UL
#define BIT16_MAX 65535UL
#define BIT8_MAX 255UL
#ifndef UCHAR_MAX
#error Size of char data type is unknown. Define UCHAR_MAX.
@ -120,6 +119,4 @@ typedef unsigned char UInt32;
#error Unable to determine suitable data type for 32-bit integers.
#endif
/* The ANSI C standard only guarantees a data size of up to 32 bits. */
#endif
#endif /* CYTOPLASM_INT_H */

249
src/include/Int64.h Normal file
View File

@ -0,0 +1,249 @@
/*
* Copyright (C) 2022-2023 Jordan Bancino <@jordan:bancino.net>
*
* Permission is hereby granted, free of charge, to any person
* obtaining a copy of this software and associated documentation files
* (the "Software"), to deal in the Software without restriction,
* including without limitation the rights to use, copy, modify, merge,
* publish, distribute, sublicense, and/or sell copies of the Software,
* and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#ifndef CYTOPLASM_INT64_H
#define CYTOPLASM_INT64_H
/***
* @Nm Int64
* @Nd Fixed-width 64 bit integers.
* @Dd August 11, 2023
*
* .Pp
* ANSI C89 (or C99 for that matter) provides no required mechanism
* for 64 bit integers. Nevertheless, many compilers provide them as
* extensions. However, since it is not a gaurantee, and to be fully
* standards-compliant and thus portable, a platform-agnostic interface
* is required. This header provides such an interface. If the platform
* has a 64 bit integer type, that is used, and native operations are
* performed by C preprocessor macro expansion. Otherwise, a
* compatibility layer is provided, which implements 64-bit
* arithmetic on an array of 2 32-bit numbers which are provided by
* .Xr Int 3 .
* .Pp
* Note that 64-bit emulation is certainly not as performant as using
* native 64-bit operations, so whenever possible, the native
* operations should be preferred. However, since C provides no required
* 64 bit integer on 32-bit and less platforms, this API can be used as
* a "good enough" fallback mechanism.
* .Pp
* Also note that this implementation, both in the native and
* non-native forms, makes some assumptions:
* .Bl -bullet -width Ds
* .It
* When a cast from a larger integer to a smaller integer is performed,
* the upper bits are truncated, not the lower bits.
* .It
* Negative numbers are represented in memory and in registers in two's
* compliment form.
* .El
* .Pp
* This API may provide unexpected output if these assumptions are
* false for a given platform.
*
* @ignore-typedefs
*/
#include <Int.h>
#include <stddef.h>
#define BIT64_MAX 18446744073709551615UL
/* TODO: Implement signed arithmetic */
#if UINT_MAX == BIT64_MAX
/* typedef signed int Int64; */
typedef unsigned int UInt64;
#define UINT64_NATIVE
#elif ULONG_MAX == BIT64_MAX
/* typedef signed int Int64; */
typedef unsigned long UInt64;
#define UINT64_NATIVE
#endif
#ifdef UINT64_NATIVE
#define UInt64Create(high, low) (((UInt64) (high) << 32) | (low))
#define UInt64Low(a) ((UInt32) ((a) & 0x00000000FFFFFFFF))
#define UInt64High(a) ((UInt32) ((a) >> 32))
#define UInt64Add(a, b) ((a) + (b))
#define UInt64Sub(a, b) ((a) - (b))
#define UInt64Mul(a, b) ((a) * (b))
#define UInt64Div(a, b) ((a) / (b))
#define UInt64Rem(a, b) ((a) % (b))
#define UInt64Bsl(a, b) ((a) << (b))
#define UInt64Bsr(a, b) ((a) >> (b))
#define UInt64And(a, b) ((a) & (b))
#define UInt64Or(a, b) ((a) | (b))
#define UInt64Xor(a, b) ((a) ^ (b))
#define UInt64Not(a) (~(a))
#define UInt64Eq(a, b) ((a) == (b))
#define UInt64Lt(a, b) ((a) < (b))
#define UInt64Gt(a, b) ((a) > (b))
#define UInt64Neq(a, b) ((a) != (b))
#define UInt64Leq(a, b) ((a) <= (b))
#define UInt64Geq(a, b) ((a) >= (b))
#else
/**
* For platforms that do not have a native integer large enough to
* store a 64 bit integer, this struct is used. i[0] contains the low
* bits of integer, and i[1] contains the high bits of the integer.
* .Pp
* This struct should not be accessed directly, because UInt64 may not
* actually be this struct, it might be an actual integer type. For
* maximum portability, only use the functions defined here to
* manipulate 64 bit integers.
*/
typedef struct
{
UInt32 i[2];
} UInt64;
/**
* Create a new unsigned 64 bit integer using the given high and low
* bits.
*/
extern UInt64 UInt64Create(UInt32, UInt32);
/**
* Add two unsigned 64 bit integers together.
*/
extern UInt64 UInt64Add(UInt64, UInt64);
/**
* Subtract the second 64 bit integer from the first.
*/
extern UInt64 UInt64Sub(UInt64, UInt64);
/**
* Multiply two 64 bit integers together. The non-native version of
* this function uses the Russian Peasant method of multiplication,
* which should afford more performance than a naive multiplication by
* addition, but it is still rather slow and depends on the size of
* the integers being multiplied.
*/
extern UInt64 UInt64Mul(UInt64, UInt64);
/**
* Divide the first 64 bit integer by the second and return the
* quotient. The non-native version of this function uses naive binary
* long division, which is slow, but gauranteed to finish in constant
* time.
*/
extern UInt64 UInt64Div(UInt64, UInt64);
/**
* Divide the first 64 bit integer by the second and return the
* remainder. The non-native version of this function uses naive binary
* long division, which is slow, but gauranteed to finish in constant
* time.
*/
extern UInt64 UInt64Rem(UInt64, UInt64);
/**
* Perform a left logical bit shift of a 64 bit integer. The second
* parameter is how many places to shift, and is declared as a regular
* integer because anything more than 64 does not make sense.
*/
extern UInt64 UInt64Sll(UInt64, int);
/**
* Perform a right logical bit shift of a 64 bit integer. The second
* parameter is how many places to shift, and is declared as a regular
* integer because anything more than 64 does not make sense.
*/
extern UInt64 UInt64Srl(UInt64, int);
/**
* Perform a bitwise AND (&) of the provided 64 bit integers.
*/
extern UInt64 UInt64And(UInt64, UInt64);
/**
* Perform a bitwise OR (|) of the provided 64 bit integers.
*/
extern UInt64 UInt64Or(UInt64, UInt64);
/**
* Perform a bitwise XOR (^) of the provided 64 bit integers.
*/
extern UInt64 UInt64Xor(UInt64, UInt64);
/**
* Perform a bitwise NOT (~) of the provided 64 bit integer.
*/
extern UInt64 UInt64Not(UInt64);
/**
* Perform a comparison of the provided 64 bit integers and return a C
* boolean that is true if and only if they are equal.
extern int UInt64Eq(UInt64, UInt64);
/**
* Perform a comparison of the provided 64 bit integers and return a C
* boolean that is true if and only if the second operand is strictly
* less than the first.
*/
extern int UInt64Lt(UInt64, UInt64);
/**
* Perform a comparison of the provided 64 bit integers and return a C
* boolean that is true if and only if the second operand is strictly
* greater than the first.
*/
extern int UInt64Gt(UInt64, UInt64);
#define UInt64Low(a) ((a).i[0])
#define UInt64High(a) ((a).i[1])
#define UInt64Neq(a, b) (!UInt64Eq(a, b))
#define UInt64Leq(a, b) (UInt64Eq(a, b) || UInt64Lt(a, b))
#define UInt64Geq(a, b) (UInt64Eq(a, b) || UInt64Gt(a, b))
#endif
#define UINT64_STRBUF 65 /* Base 2 representation with '\0' */
/**
* Convert a 64 bit integer to a string in an arbitrary base
* representation specified by the second parameter, using the provided
* buffer and length specified by the third and fourth parameters. To
* guarantee that the string will fit in the buffer, allocate it of
* size UINT64_STRBUF or larger. Note that a buffer size smaller than
* UINT64_STRBUF will invoke undefined behavior.
*/
extern size_t UInt64Str(UInt64, int, char *, size_t);
#endif /* CYTOPLASM_INT64_H */

117
tools/int64.c Normal file
View File

@ -0,0 +1,117 @@
#include <Int64.h>
#include <Log.h>
/* AssertEquals(actual, expected) */
int AssertEquals(char *msg, UInt64 x, UInt64 y)
{
if (!UInt64Eq(x, y))
{
Log(LOG_ERR, "%s: Expected 0x%X 0x%X, got 0x%X 0x%X", msg,
UInt64High(y), UInt64Low(y),
UInt64High(x), UInt64Low(x));
return 0;
}
return 1;
}
int Main(void)
{
UInt64 x, y;
Log(LOG_INFO, "sizeof(UInt64) = %lu", sizeof(UInt64));
#ifdef UINT64_NATIVE
Log(LOG_INFO, "Using native 64-bit integers.");
#else
Log(LOG_INFO, "Using emulated 64-bit integers.");
#endif
/* BSR Tests */
x = UInt64Create(0x000000FF, 0x00000000);
y = UInt64Bsr(x, 4);
AssertEquals("x >> 4", y, UInt64Create(0x0000000F, 0xF0000000));
y = UInt64Bsr(x, 8);
AssertEquals("x >> 8", y, UInt64Create(0x00000000, 0xFF000000));
y = UInt64Bsr(x, 36);
AssertEquals("x >> 36", y, UInt64Create(0x00000000, 0x0000000F));
/* BSL Tests */
x = UInt64Create(0x00000000, 0xFF000000);
y = UInt64Bsl(x, 4);
AssertEquals("x << 4", y, UInt64Create(0x0000000F, 0xF0000000));
y = UInt64Bsl(x, 8);
AssertEquals("x << 8", y, UInt64Create(0x000000FF, 0x00000000));
y = UInt64Bsl(x, 36);
AssertEquals("x << 36", y, UInt64Create(0xF0000000, 0x00000000));
/* ADD Tests */
x = UInt64Create(0x00000000, 0xF0000001);
y = UInt64Create(0x00000000, 0x00000002);
AssertEquals("0xF0000001 + 0x00000002", UInt64Add(x, y), UInt64Create(0x00000000, 0xF0000003));
x = UInt64Create(0x00000000, 0xF0000000);
y = UInt64Create(0x00000000, 0x10000000);
AssertEquals("0xF0000000 + 0x10000000", UInt64Add(x, y), UInt64Create(0x00000001, 0x00000000));
/* SUB Tests */
x = UInt64Create(0x00000000, 0x00000005);
y = UInt64Create(0x00000000, 0x00000002);
AssertEquals("0x00000005 - 0x00000002", UInt64Sub(x, y), UInt64Create(0x00000000, 0x00000003));
x = UInt64Create(0x00000001, 0x00000000);
y = UInt64Create(0x00000000, 0x00000001);
AssertEquals("0x00000001 0x00000000 - 0x00000001", UInt64Sub(x, y), UInt64Create(0x00000000, 0xFFFFFFFF));
/* MUL Tests */
x = UInt64Create(0, 18);
y = UInt64Create(0, 1);
AssertEquals("18 * 1", UInt64Mul(x, y), UInt64Create(0, 18));
x = UInt64Create(0, 20);
y = UInt64Create(0, 12);
AssertEquals("20 * 12", UInt64Mul(x, y), UInt64Create(0, 240));
x = UInt64Create(0x00000000, 0x00000005);
y = UInt64Create(0x00000000, 0x00000005);
AssertEquals("0x00000005 * 0x00000005", UInt64Mul(x, y), UInt64Create(0x00000000, 0x00000019));
x = UInt64Create(0x00000001, 0x00000000);
y = UInt64Create(0x00000000, 0x00000005);
AssertEquals("0x00000001 0x00000000 * 0x00000005", UInt64Mul(x, y), UInt64Create(0x00000005, 0x00000000));
/* DIV Tests */
x = UInt64Create(0, 12);
y = UInt64Create(0, 4);
AssertEquals("12 / 4", UInt64Div(x, y), UInt64Create(0, 3));
/* MOD Tests */
x = UInt64Create(0x000000FF, 0x00000000);
y = UInt64Create(0x00000000, 0x00000010);
AssertEquals("0x000000FF 0x00000000 mod 0x00000010", UInt64Rem(x, y), UInt64Create(0, 0));
x = UInt64Create(0x00000000, 0xFF000000);
y = UInt64Create(0x00000000, 0x00000010);
AssertEquals("0x00000000 0xFF000000 mod 0x00000010", UInt64Rem(x, y), UInt64Create(0, 0));
x = UInt64Create(0xFF000000, 0x00000000);
y = UInt64Create(0x00000000, 0x00000010);
AssertEquals("0xFF000000 0x00000000 mod 0x00000010", UInt64Rem(x, y), UInt64Create(0, 0));
x = UInt64Create(0x00000000, 0x000000F0);
y = UInt64Create(0x00000000, 0x00000010);
AssertEquals("0x00000000 0x000000F0 mod 0x00000010", UInt64Rem(x, y), UInt64Create(0, 0));
return 0;
}