Comparing elapsed time to copy memory buffers: memcpy vs ForLoopIndexCopying - c++

Question

I am trying to understand the difference between "memcpy" and "for-loop-index-copying" of a buffer in the time it takes to copy.
results:
CopyingType | MemoryType | RunMode | Elapsed Time(ms)
-----------------------------------------------------------
memcpy | stack | Debug | x
forLoopIndexing | stack | Debug | 30x
memcpy | stack | Release | 0
forLoopIndexing | stack | Release | 0
memcpy | heap | Release | 0
forLoopIndexing | heap | Release | 2000
Here is my code to run... Maybe I am doing something wrong ???
Seems odd that to copy a 500,000 byte buffer 100,000 times takes no time at all or at least less than the resolution of the machine... in my case 16ms.
#include "stdafx.h"
#include <windows.h>
#include <iostream>
int main()
{
long baseTime;
const long packetLength = 500000;
//char packet1[packetLength];//stack
//char packet2[packetLength];//stack
char *packet1 = (char*)calloc(packetLength, sizeof(char));//heap
char *packet2 = (char*)calloc(packetLength, sizeof(char));//heap
memset(packet1, 0, packetLength);//init
memset(packet2, 0, packetLength);//init
long NumPackets = 100000;
long NumRuns = 10;
for (long k = 0; k < NumRuns; k++)
{
//create packet
printf("\npacket1:\n");
for (long i = 0; i < packetLength; i++) {
packet1[i] = (char)(i % 26 + 65);
}
printf("\nk:%d\n", k);
//index copy
baseTime = GetTickCount();
long ii = 0;
for (long j = 0; j < NumPackets; j++) {
for (long i = 0; i < packetLength; i++) {
packet2[i] = packet1[i];
}
}
printf("Time(IndexCopy): %ld\n", GetTickCount() - baseTime);
//memcpy
memset(packet2, 0, packetLength);//reset
baseTime = GetTickCount();
for (long j = 0; j < NumPackets; j++) {
memcpy(packet2, packet1, packetLength); //Changed via PaulMcKenzie.
}
printf("Time(memcpy): %ld\n", GetTickCount() - baseTime);
//printf("\npacket2\n");
for (long i = 0; i < packetLength; i++) {
//printf("%c", packet2[i]);
}
}
int iHalt;
scanf_s("%d", &iHalt);
return 0;
}
Via the change... the new table
CopyingType | MemoryType | RunMode | Elapsed Time(ms)
-----------------------------------------------------------
memcpy | stack | Debug | x
forLoopIndexing | stack | Debug | 50x
memcpy | stack | Release | 0
forLoopIndexing | stack | Release | 0
memcpy | heap | Release | 2000
forLoopIndexing | heap | Release | 2000

Answer 1

Maybe I am doing something wrong ???
You are doing something wrong, more to the point, in the code that uses memcpy.
const long packetLength = 500000;
char *packet1 = (char*)calloc(packetLength, sizeof(char));
char *packet2 = (char*)calloc(packetLength, sizeof(char));
//...
for (long j = 0; j < NumPackets; j++) {
memcpy(packet2, packet1, sizeof(packet2)); // <-- Incorrect
}
The sizeof(packet2) is the same as sizeof(char *), which more than likely is either 4 or 8.
What you want is not the sizeof(char *), but the actual number of bytes to copy.
for (long j = 0; j < NumPackets; j++) {
memcpy(packet2, packet1, packetLength);
}