代表数字7的8位像这样:
00000111
设置了三个比特。
确定32位整数中设置位数的算法是什么?
代表数字7的8位像这样:
00000111
设置了三个比特。
确定32位整数中设置位数的算法是什么?
当前回答
另一个汉明权重算法,如果你使用的是BMI2 CPU:
the_weight = __tzcnt_u64(~_pext_u64(data[i], data[i]));
其他回答
我给出了两个算法来回答这个问题,
package countSetBitsInAnInteger;
import java.util.Scanner;
public class UsingLoop {
public static void main(String[] args) {
Scanner in = new Scanner(System.in);
try {
System.out.println("Enter a integer number to check for set bits in it");
int n = in.nextInt();
System.out.println("Using while loop, we get the number of set bits as: " + usingLoop(n));
System.out.println("Using Brain Kernighan's Algorithm, we get the number of set bits as: " + usingBrainKernighan(n));
System.out.println("Using ");
}
finally {
in.close();
}
}
private static int usingBrainKernighan(int n) {
int count = 0;
while(n > 0) {
n& = (n-1);
count++;
}
return count;
}
/*
Analysis:
Time complexity = O(lgn)
Space complexity = O(1)
*/
private static int usingLoop(int n) {
int count = 0;
for(int i=0; i<32; i++) {
if((n&(1 << i)) != 0)
count++;
}
return count;
}
/*
Analysis:
Time Complexity = O(32) // Maybe the complexity is O(lgn)
Space Complexity = O(1)
*/
}
这不是最快或最好的解决方案,但我以自己的方式发现了同样的问题,我开始反复思考。最后我意识到它可以这样做,如果你从数学方面得到这个问题,画一个图,然后你发现它是一个有周期部分的函数,然后你意识到周期之间的差异……所以你看:
unsigned int f(unsigned int x)
{
switch (x) {
case 0:
return 0;
case 1:
return 1;
case 2:
return 1;
case 3:
return 2;
default:
return f(x/4) + f(x%4);
}
}
这是一个可移植的模块(ANSI-C),它可以在任何架构上对每个算法进行基准测试。
你的CPU有9位字节?目前它实现了2个算法,K&R算法和一个字节查找表。查找表的平均速度比K&R算法快3倍。如果有人能想出办法使“黑客的喜悦”算法可移植,请随意添加它。
#ifndef _BITCOUNT_H_
#define _BITCOUNT_H_
/* Return the Hamming Wieght of val, i.e. the number of 'on' bits. */
int bitcount( unsigned int );
/* List of available bitcount algorithms.
* onTheFly: Calculate the bitcount on demand.
*
* lookupTalbe: Uses a small lookup table to determine the bitcount. This
* method is on average 3 times as fast as onTheFly, but incurs a small
* upfront cost to initialize the lookup table on the first call.
*
* strategyCount is just a placeholder.
*/
enum strategy { onTheFly, lookupTable, strategyCount };
/* String represenations of the algorithm names */
extern const char *strategyNames[];
/* Choose which bitcount algorithm to use. */
void setStrategy( enum strategy );
#endif
.
#include <limits.h>
#include "bitcount.h"
/* The number of entries needed in the table is equal to the number of unique
* values a char can represent which is always UCHAR_MAX + 1*/
static unsigned char _bitCountTable[UCHAR_MAX + 1];
static unsigned int _lookupTableInitialized = 0;
static int _defaultBitCount( unsigned int val ) {
int count;
/* Starting with:
* 1100 - 1 == 1011, 1100 & 1011 == 1000
* 1000 - 1 == 0111, 1000 & 0111 == 0000
*/
for ( count = 0; val; ++count )
val &= val - 1;
return count;
}
/* Looks up each byte of the integer in a lookup table.
*
* The first time the function is called it initializes the lookup table.
*/
static int _tableBitCount( unsigned int val ) {
int bCount = 0;
if ( !_lookupTableInitialized ) {
unsigned int i;
for ( i = 0; i != UCHAR_MAX + 1; ++i )
_bitCountTable[i] =
( unsigned char )_defaultBitCount( i );
_lookupTableInitialized = 1;
}
for ( ; val; val >>= CHAR_BIT )
bCount += _bitCountTable[val & UCHAR_MAX];
return bCount;
}
static int ( *_bitcount ) ( unsigned int ) = _defaultBitCount;
const char *strategyNames[] = { "onTheFly", "lookupTable" };
void setStrategy( enum strategy s ) {
switch ( s ) {
case onTheFly:
_bitcount = _defaultBitCount;
break;
case lookupTable:
_bitcount = _tableBitCount;
break;
case strategyCount:
break;
}
}
/* Just a forwarding function which will call whichever version of the
* algorithm has been selected by the client
*/
int bitcount( unsigned int val ) {
return _bitcount( val );
}
#ifdef _BITCOUNT_EXE_
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
/* Use the same sequence of pseudo random numbers to benmark each Hamming
* Weight algorithm.
*/
void benchmark( int reps ) {
clock_t start, stop;
int i, j;
static const int iterations = 1000000;
for ( j = 0; j != strategyCount; ++j ) {
setStrategy( j );
srand( 257 );
start = clock( );
for ( i = 0; i != reps * iterations; ++i )
bitcount( rand( ) );
stop = clock( );
printf
( "\n\t%d psudoe-random integers using %s: %f seconds\n\n",
reps * iterations, strategyNames[j],
( double )( stop - start ) / CLOCKS_PER_SEC );
}
}
int main( void ) {
int option;
while ( 1 ) {
printf( "Menu Options\n"
"\t1.\tPrint the Hamming Weight of an Integer\n"
"\t2.\tBenchmark Hamming Weight implementations\n"
"\t3.\tExit ( or cntl-d )\n\n\t" );
if ( scanf( "%d", &option ) == EOF )
break;
switch ( option ) {
case 1:
printf( "Please enter the integer: " );
if ( scanf( "%d", &option ) != EOF )
printf
( "The Hamming Weight of %d ( 0x%X ) is %d\n\n",
option, option, bitcount( option ) );
break;
case 2:
printf
( "Please select number of reps ( in millions ): " );
if ( scanf( "%d", &option ) != EOF )
benchmark( option );
break;
case 3:
goto EXIT;
break;
default:
printf( "Invalid option\n" );
}
}
EXIT:
printf( "\n" );
return 0;
}
#endif
对于232查找表和逐个遍历每个位之间的折中方法:
int bitcount(unsigned int num){
int count = 0;
static int nibblebits[] =
{0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
for(; num != 0; num >>= 4)
count += nibblebits[num & 0x0f];
return count;
}
从http://ctips.pbwiki.com/CountBits
我觉得很无聊,于是对三种方法进行了十亿次迭代。编译器是gcc -O3。CPU就是第一代Macbook Pro里装的东西。
最快的是3.7秒:
static unsigned char wordbits[65536] = { bitcounts of ints between 0 and 65535 };
static int popcount( unsigned int i )
{
return( wordbits[i&0xFFFF] + wordbits[i>>16] );
}
第二名是相同的代码,但查找的是4个字节而不是2个半字。这花了大约5.5秒。
第三名是“横向加法”法,用时8.6秒。
第四名是GCC的__builtin_popcount(),仅为11秒。
一次一个比特的计数方法要慢得多,我厌倦了等待它完成。
因此,如果您最关心的是性能,那么请使用第一种方法。如果您关心它,但又不想在上面花费64Kb的RAM,那么可以使用第二种方法。否则,请使用可读的(但速度较慢)一次一位的方法。
很难想象在什么情况下你会想要使用比特旋转方法。
编辑:这里也有类似的结果。