What is being cached?

I am trying to time a chunk of code. If I run it from the command line like: Code: taskset -c 0 ./conv Conv Time:  0.014350082 I get a time of ~14ms but there is some variation. So, I put the command above into a script and ran to get an average Code: taskset -c 0 ./conv taskset -c 0 ./conv . . . taskset -c 0 ./conv The output I get from running the script is: Code: Conv Time:  0.014284663 Conv Time:  0.010995277 Conv Time:  0.008977601 Conv Time:  0.008802842 Conv Time:  0.009022121 Conv Time:  0.008894075 Conv Time:  0.008738924 Conv Time:  0.008423263 Conv Time:  0.008694661 Conv Time:  0.008902102 What is causing the increased performance I see? If the program is starting and exiting, what could be cached? I also ran another version that I moved the program to a new core each time in the script. Code: taskset -c 0 ./conv taskset -c 1 ./conv . . . taskset -c 9 ./conv That gave an output of: Code: Conv Time:  0.014503283 Conv Time:  0.013998719 Conv Time:  0.013912381 Conv Time:  0.014000417 Conv Time:  0.013930824 Conv Time:  0.013995235 Conv Time:  0.014705306 Conv Time:  0.013931455 Conv Time:  0.014027264 Conv Time:  0.014017024 Give that, something is helping the performance if I say on the same core. Here is the code below Code: // Use: //        g++ -o conv -O3 conv.cpp -lrt // to compile #include <iostream> #include <cstdio> #include <sys/time.h> using namespace std; typedef struct {     float real;     float imag; } Complex_T; #define TAPS  228 #define SAMPS 341 int main(int argc, char **argv) {   Complex_T y[TAPS+SAMPS],y2[TAPS+SAMPS],h[TAPS],x[SAMPS];   int i(0),j(0);   struct timeval val1,val2;   struct timespec a,b;   double thetime1,thetime2;   clock_gettime(CLOCK_REALTIME,&a);   for (int k(0); k < 100; k++)   for ( i = 0; i < TAPS+SAMPS; i++ ) {     y[i].real = 0;                      // set to zero before sum     y[i].imag = 0;                      // set to zero before sum     for ( j = 0; j < TAPS; j++ ){         if ( i-j < 0 )           break;         if ( i-j >= SAMPS )           continue;         y[i].real += h[j].real * x[i - j].real;    // convolve: multiply and accumulate         y[i].imag += h[j].real * x[i - j].imag;    // convolve: multiply and accumulate         y[i].imag += h[j].imag * x[i - j].real;    // convolve: multiply and accumulate         y[i].real += -1.0*(h[j].imag * x[i - j].imag);    // convolve: multiply and accumulate     }   }   clock_gettime(CLOCK_REALTIME,&b);   thetime1 = (double)((double)a.tv_sec+(double)(a.tv_nsec)/1.0e9);   thetime2 = (double)((double)b.tv_sec+(double)(b.tv_nsec)/1.0e9);   printf("Conv Time:  %.9f\n",(thetime2-thetime1));   return 0; }

URL: