Micro benchmark memmove vs memcpy

microbench.cc

// compile with
// g++ -Wall microbench.cc -lboost_system -lboost_chrono -Ofast -o microbench
//
#include <string.h>

#include <iostream>
#include <boost/chrono.hpp>
#include <iostream>
#include <iomanip>

using namespace boost::chrono;

// based on
// http://www.boost.org/doc/libs/1_54_0/libs/chrono/example/await_keystroke.cpp
template<class Clock>
class Timer
{
    typename Clock::time_point start;

  public:
    Timer()
      : start(Clock::now()) {
    }

    typename Clock::duration elapsed() const {
      return (Clock::now() - start);
    }

    double seconds() const {
      return (elapsed().count() *
            ((double)Clock::period::num / Clock::period::den));
    }
};

#define BUFFERSIZE  (100 * 1024 * 1024)  // 100 mb
#define LOOPS       10
#define MAXSTRIDE   128

int main()
{
  char *b1 = (char *)::malloc(BUFFERSIZE + MAXSTRIDE);
  char *b2 = (char *)::malloc(BUFFERSIZE + MAXSTRIDE);
  ::memset(b1, 0, BUFFERSIZE + MAXSTRIDE);

  // Each test is performed |LOOP| times; the average run-time is then printed.

  // Test memcpy() by copying from buffer b1 to b2
  Timer<boost::chrono::high_resolution_clock> t;
  for (int i = 0; i < LOOPS; i++)
    memcpy(b2, b1, BUFFERSIZE);
  double seconds = t.seconds();
  std::cout << "memcpy        " << (seconds / LOOPS) << std::endl;

  // Test memmove() by inserting a small gap at the beginning of the buffer
  for (int stride = 2; stride <= MAXSTRIDE; stride *= 2) {
    Timer<boost::chrono::high_resolution_clock> t;
    for (int i = 0; i < LOOPS; i++)
      memmove(b1 + stride, b1, BUFFERSIZE);
    double seconds = t.seconds();
    std::cout << "memmove (" << std::setfill('0') << std::setw(3)
            << stride << ") " << (seconds / LOOPS) << std::endl;
  }

  ::free(b1);
  ::free(b2);
  return (0);
}