Access speed of heap / stack / program data area variables?

xianzeming

In the era of dos, there are near pointers and far pointers in programs. Near pointers only need to address offset addresses, which is faster. Far pointers also need to cooperate with segment registers when accessing memory.
But for Win32 32-bit flat mode programming, the memory access does not require the cooperation of the segment register or the segment register has changed its use, then the memory access of the stack, heap and global variables in the data area is pure Is there still a difference in access speed? Have you done any experiments in this area?

永恒之骑士

Learn. . . Look at the expert opinions downstairs

paladinkan

attention

lcdxhj

There is basically no difference ...

// DumpStack.cpp: Defines the entry point for the console application.
//

#include "stdafx.h"
#include <windows.h>

#define MAX_NUMBER_OF_DATA 1000

int g_dataArray [MAX_NUMBER_OF_DATA];

void global ()
{
// Testing ...
LARGE_INTEGER c1, c2;
:: QueryPerformanceCounter (&c1);
for (int i = 0; i <MAX_NUMBER_OF_DATA; ++ i)
{
g_dataArray [i] = i;
}
:: QueryPerformanceCounter (&c2);
printf ("### Global performance count:% I64u\n", c2.QuadPart-c1.QuadPart);
}

void stack ()
{
int dataArray [MAX_NUMBER_OF_DATA];
// Testing ...
LARGE_INTEGER c1, c2;
:: QueryPerformanceCounter (&c1);
for (int i = 0; i <MAX_NUMBER_OF_DATA; ++ i)
{
dataArray [i] = i;
}
:: QueryPerformanceCounter (&c2);
printf ("### Stacking performance count:% I64u\n", c2.QuadPart-c1.QuadPart);
}

void dump ()
{
int * pDataArray = new int [MAX_NUMBER_OF_DATA];
// Testing ...
LARGE_INTEGER c1, c2;
:: QueryPerformanceCounter (&c1);
for (int i = 0; i <MAX_NUMBER_OF_DATA; ++ i)
{
pDataArray [i] = i;
}
:: QueryPerformanceCounter (&c2);
printf ("### Dumping performance count:% I64u\n", c2.QuadPart-c1.QuadPart);
delete [] pDataArray;
pDataArray = NULL;
}

int main (int argc, char * argv [])
{
for (int i = 0; i <100; ++ i)
{
printf ("======= No.% d ======\n", i);
global ();
dump ();
stack ();
}

return 0;
}

ilovevc2019

Follow and learn.

xianzeming

In fact, there is still a certain gap if the amount of data is magnified another 1,000 times.