注意:正如 @Damien_The_Unbeliever 在 cmets 中指出的那样,upperHigh 和 lowerLow 表是相同的。所以它们可以组合成一个表。但是,这种优化会使代码更难阅读,解释也更难写,因此将表格组合起来留给读者作为练习。
下面的代码显示了如何在将 16 位无符号值除以 7 时生成商和余数。解释代码 (IMO) 的最简单方法是举例说明,所以让我们考虑将0xa732 除以 7。预期结果是:
quotient = 0x17e2
remainder = 4
我们首先将输入视为两个 8 位值,upper 字节和 lower 字节。 upper 字节为0xa7,lower 字节为0x32。
我们从 upper 字节计算商和余数:
0xa700 / 7 = 0x17db
0xa700 % 7 = 3
所以我们需要三个表:
-
upperHigh存储商的高字节:upperHigh[0xa7] = 0x17
-
upperLow 存储商的低字节:upperLow[0xa7] = 0xdb
-
upperRem 存储剩余部分:upperRem[0xa7] = 3
我们计算lower字节的商和余数:
0x32 / 7 = 0x07
0x32 % 7 = 1
所以我们需要两张表:
-
lowerLow 存储商的低字节:lowerLow[0x32] = 0x07
-
lowerRem 存储余数:lowerRem[0x32] = 1
现在我们需要汇总最终答案。余数是两个余数之和。由于每个余数都在 [0,6] 范围内,因此总和在 [0,12] 范围内。因此,我们可以使用两个 13 字节查找将和转换为最终余数和进位。
商的低字节是该进位与lowerLow 和upperLow 表中的值之和。请注意,总和可能会在高字节中产生进位。
商的高字节是该进位与upperHigh 表中的值之和。
所以,完成这个例子:
remainder = 1 + 3 = 4 // simple add (no carry in)
lowResult = 0x07 + 0xdb = 0xe2 // add with carry from remainder
highResult = 0x17 // add with carry from lowResult
实现这一点的汇编代码由 7 个表查找、一个不带进位的加法指令和两个带进位的加法指令组成。
#include <stdio.h>
#include <stdint.h>
uint8_t upperHigh[256]; // index:(upper 8 bits of the number) value:(high 8 bits of the quotient)
uint8_t upperLow[256]; // index:(upper 8 bits of the number) value:(low 8 bits of the quotient)
uint8_t upperRem[256]; // index:(upper 8 bits of the number) value:(remainder when dividing the upper bits by 7)
uint8_t lowerLow[256]; // index:(lower 8 bits of the number) value:(low 8 bits of the quotient)
uint8_t lowerRem[256]; // index:(lower 8 bits of the number) value:(remainder when dividing the lower bits by 7)
uint8_t carryRem[13] = { 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1 };
uint8_t combinedRem[13] = { 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5 };
void populateLookupTables(void)
{
for (uint16_t i = 0; i < 256; i++)
{
uint16_t upper = i << 8;
upperHigh[i] = (upper / 7) >> 8;
upperLow[i] = (upper / 7) & 0xff;
upperRem[i] = upper % 7;
uint16_t lower = i;
lowerLow[i] = lower / 7;
lowerRem[i] = lower % 7;
}
}
void divideBy7(uint8_t upperValue, uint8_t lowerValue, uint8_t *highResult, uint8_t *lowResult, uint8_t *remainder)
{
uint8_t temp = upperRem[upperValue] + lowerRem[lowerValue];
*remainder = combinedRem[temp];
*lowResult = upperLow[upperValue] + lowerLow[lowerValue] + carryRem[temp];
uint8_t carry = (upperLow[upperValue] + lowerLow[lowerValue] + carryRem[temp]) >> 8; // Note this is just the carry flag from the 'lowResult' calcaluation
*highResult = upperHigh[upperValue] + carry;
}
int main(void)
{
populateLookupTables();
uint16_t n = 0;
while (1)
{
uint8_t upper = n >> 8;
uint8_t lower = n & 0xff;
uint16_t quotient1 = n / 7;
uint16_t remainder1 = n % 7;
uint8_t high, low, rem;
divideBy7(upper, lower, &high, &low, &rem);
uint16_t quotient2 = (high << 8) | low;
uint16_t remainder2 = rem;
printf("n=%u q1=%u r1=%u q2=%u r2=%u", n, quotient1, remainder1, quotient2, remainder2);
if (quotient1 != quotient2 || remainder1 != remainder2)
printf(" **** failed ****");
printf("\n");
n++;
if (n == 0)
break;
}
}