Thread: project Euler problem 5 solution

Originally Posted by Click_here

Code:

#include "cycle_counting.h"

Okay... You can keep your secretes :P

No secret at all... nobody, yet, had the curiosity to ask me about this. Here's the code to measure clocks on SandyBridge and above processors (Intel) and most AMD ones, as well on ARM AArch64:

Code:

/* cycle_counting.h */
#ifndef CYCLE_COUNTING_INCLUDED__
#define CYCLE_COUNTING_INCLUDED__

#ifndef __GNUC__
# error Works only on GCC
#endif

/* ==========================================
    Quick & Dirty cycle counting...

    Usage:

      counter_T cnt;

      cnt = BEGIN_TSC();      
      f();
      END_TSC(&cnt);

    For i386 and x86-64 you can define SYNC_MEM if want a better
    processor serialization before measuring the cycles (requires SSE2).
   ========================================== */
#include <stdint.h>

// Non volatitle to avoid optimizations.
typedef volatile uint64_t counter_T;

#if defined(__x86_64__)
  inline counter_T BEGIN_TSC( void )
  {
    uint32_t a, d;

    __asm__ __volatile__ (
# ifdef SYNC_MEM
      "mfence\n\t"
# endif
      "xorl %%eax,%%eax\n\t"
      "cpuid\n\t"
      "rdtsc" : "=a" (a), "=d" (d) :: "%rbx", "%rcx"
    );

    return a | ((uint64_t)d << 32);
  }

  inline void END_TSC( counter_T *cptr )
  {
    uint32_t a, d;

    __asm__ __volatile__ (
      "rdtscp" : "=a" (a), "=d" (d) :: "%rcx"
    );

    *cptr = (a | ((uint64_t)d << 32)) - *cptr;;
  }
#elif defined(__i386__)
  inline counter_T BEGIN_TSC( void )
  {
    uint32_t a, d;

    __asm__ __volatile__ (
// mfence available only to SSE2 and above.
# ifdef __SSE2__
#   ifdef SYNC_MEM
      "mfence\n\t"
#   endif
# endif
      "xorl %%eax,%%eax\n\t"
      "cpuid\n\t"
      "rdtsc" : "=a" (a), "=d" (d) :: "%ebx", "%ecx"
    );

    return a | ((uint64_t)d << 32);
  }

  inline void END_TSC( counter_T *cptr )
  {
    uint32_t a, d;

    __asm__ __volatile__ (
      "rdtscp" : "=a" (a), "=d" (d) :: "%ecx"
    );

    *cptr = (a | ((uint64_t)d << 32)) - *cptr;
  }
#elif defined(__aarch64__)
  inline counter_T BEGIN_TSC( void )
  {
    uint64_t count;

    __asm__ __volatile__ ( 
# ifdef SYNC_MEM
    "dmb\n\t"
# endif
    "mrs %0,cntvct_el0" : "=r" (count) );

    return count;
  }

  inline void END_TSC( counter_T *cptr )
  {
    uint64_t count;

    __asm__ __volatile__ ( "mrs %0,cntvct_el0" : "=r" (count) );

    *cptr = count - *cptr;
  }
#else
# error i386, x86-64 and AArch64 only.
#endif

#endif

Originally Posted by Click_here

@flp - "\033[31;" does not work well on Windows :P

WTF is "Windows"?

Originally Posted by laserlight

Noun. An opening in the wall or roof of a building or vehicle, fitted with glass in a frame to admit light or air and allow people to see out.

Hehehehe...

Originally Posted by flp1969

WTF is "Windows"?

The aperture through which Microsoft invades your privacy and attempts to control all you see and do, with partnerships invited through the Internet to observe and influence what you buy.

ok those last few comments made me spit choke on my cuppa lol

@ flp did you try running the for loop from max down to i=3 if so did it make any time difference