Memory check - Mtrace vs Valgrind (how long does it take)

Mtrace vs Valgrind

mtrace와 valgrind 둘 중 어느것이 memory 오류 검출 시 걸리는 시간을 비교했습니다.

테스트 code 1.

(mtrace 사용 시 주석 해제 후 재빌드)

#include <stdio.h>
#include <stdlib.h>
//#include <mcheck.h>

int main()
{
 char *test;

 for(int i=0; i<10000; i++ )
 {
  //mtrace();
  test = malloc(10);
  free(test);
  //muntrace();
 }
 printf("hello world\n");
 return 0;
}

속도 비교

일반 실행

-bash-4.2$ time ./hello
hello world

real    0m0.002s
user    0m0.000s
sys     0m0.002s

mtrace 사용 시

-bash-4.2$ time ./hello
hello world

real    0m0.234s
user    0m0.039s
sys     0m0.188s

valgrind 사용 시

-bash-4.2$ time valgrind --leak-check=full ./hello
==15079== Memcheck, a memory error detector
==15079== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==15079== Using Valgrind-3.9.0 and LibVEX; rerun with -h for copyright info
==15079== Command: ./hello
==15079==
hello world
==15079==
==15079== HEAP SUMMARY:
==15079==     in use at exit: 0 bytes in 0 blocks
==15079==   total heap usage: 10,000 allocs, 10,000 frees, 100,000 bytes allocated
==15079==
==15079== All heap blocks were freed -- no leaks are possible
==15079==
==15079== For counts of detected and suppressed errors, rerun with: -v
==15079== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 2 from 2)

real    0m0.335s
user    0m0.314s
sys     0m0.021s

테스트 코드2.

(mtrace 사용 시 주석 해제 후 재빌드)

#include <stdio.h>
#include <stdlib.h>
//#include <mcheck.h>

int main()
{
 char *test;

 for(int i=0; i<10000; i++ )
 {
  for(int j=0; j<100; j++)
  {
   //mtrace();
   test = malloc(10);
   free(test);
   //muntrace();
  }
 }
 printf("hello world\n");
 return 0;
}

속도 비교

일반 실행

-bash-4.2$ time ./hello
hello world

real    0m0.029s
user    0m0.029s
sys     0m0.001s

mtrace 사용 시

-bash-4.2$ time ./hello
hello world

real    0m27.613s
user    0m3.842s
sys     0m23.636s

valginrd 사용 시

-bash-4.2$ time valgrind ./hello
==24091== Memcheck, a memory error detector
==24091== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==24091== Using Valgrind-3.9.0 and LibVEX; rerun with -h for copyright info
==24091== Command: ./hello
==24091==
hello world
==24091==
==24091== HEAP SUMMARY:
==24091==     in use at exit: 0 bytes in 0 blocks
==24091==   total heap usage: 1,000,000 allocs, 1,000,000 frees, 10,000,000 bytes allocated
==24091==
==24091== All heap blocks were freed -- no leaks are possible
==24091==
==24091== For counts of detected and suppressed errors, rerun with: -v
==24091== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 2 from 2)

real    0m3.124s
user    0m3.023s
sys     0m0.094s

결론

mtrace가 code상에 들어가서 빠를것이라는 예상과 다르게 코드 복잡도가 올라갈수록 valgrind가 빠르게 동작했다.