int main(int argc, char *argv[]) {
int row1,col1,row2;
//============================================================================
// Initialize array elements
for(row1=0; row1<ARRAY_SIZE; row1++)
{
for(col1=0;col1<ARRAY_SIZE;col1++)
{
arrayInPPE1[row1][col1] = rand()%100;
}
}
long unsigned int start_t,end_t;
start_t = __mftb();
//=========================================================================
/*Loop for array computation*/
for(row1 = 0;row1<ARRAY_SIZE;row1++)
{
for(col1=0;col1< ARRAY_SIZE;col1++)
{
arrayResult[row1][col1] = 0;
for(row2 = 0;row2< ARRAY_SIZE;row2++)
{
arrayResult[row1][col1] += arrayInPPE1[row1][row2]*arrayInPPE1[row2][col1];
}
}
}
end_t = __mftb();
//=========================================================================
/* Print out the multiplication result */
for(row1=0; row1<ARRAY_SIZE; row1++)
{
printf("\n");
for(col1=0;col1<ARRAY_SIZE;col1++)
{
printf("%d\t",arrayResult[row1][col1]);
}
}
printf("\nArray multiplication computation time is %f us\n",(float)(end_t-start_t)/79.8);
return 0;
}
unsigned long long _bgq_GetTimeBase( void )
{
#if defined (__GNUC__)
return GetTimeBase();
#elif defined (__IBMC__)
return __mftb();
#else
#error "Cannot find GetTimeBase for BG/Q (unhandled compiler)"
#endif
}
static inline double
ppc_intrinsic_time (void)
{
unsigned long hi, lo;
do
{
hi = __mftbu();
lo = __mftb();
} while (hi != (unsigned long) __mftbu());
return (hi * 0x100000000ull + lo) * timeBaseRatio;
}
int main(int argc, char *argv[]) {
int rc;
unsigned int i,j,l,m;
//============================================================================
// Initialize array elements
for(i=0; i<ARRAY_SIZE; i++)
{
for(j=0;j<ARRAY_SIZE;j++)
{
Matrix[i][j] = rand()%1000;
Transpose[j][i]=Matrix[i][j];
}
}
printf("Computing... Wait for some time!!!\n");
#ifdef TIMING
/********************************************************/
long unsigned int start_t, end_t;
start_t=__mftb();
/********************************************************/
#endif
// Fill in control block
//============================================================================
// create SPE context and load SPE program into the SPE context
for(i=0;i<64;i++)
{
for(j=0;j<8;j++)
{
for(l=0;l<8;l++)
{
//printf("\n%d %d %d",i,j,l);
cb[l].data=(unsigned int)Matrix[i*16];
cb[l].data1=(unsigned int)Transpose[((j*8)+l)*16];
cb[l].result=(unsigned int)&Mult[0][(i*64+j*8+l)*256];
if ((data[l].speid = spe_context_create (0, NULL)) == NULL)
{
fprintf (stderr, "Failed spe_context_create(errno=%d strerror=%s)\n", errno, strerror(errno));
exit (3);
}
//Load program
if ((rc = spe_program_load (data[l].speid, &spu)) != 0)
{
fprintf (stderr, "Failed spe_program_load(errno=%d strerror=%s)\n", errno, strerror(errno));
exit (3);
}
data[l].argp = (unsigned long long *) &cb[l];
// create SPE pthreads
if ((rc = pthread_create (&data[l].pthread, NULL, &ppu_pthread_function, &data[l])) != 0)
{
fprintf (stderr, "Failed pthread_create(errno=%d strerror=%s)\n", errno, strerror(errno));
exit (3);
}
}
// wait for SPE to complete
for(m=0;m<8;m++)
{
if ((rc = pthread_join (data[m].pthread, NULL)) != 0)
{
fprintf (stderr, "Failed pthread_join(rc=%d, errno=%d strerror=%s)\n", rc, errno, strerror(errno));
exit (1);
}
if ((rc = spe_context_destroy (data[m].speid)) != 0)
{
fprintf (stderr, "Failed spe_context_destroy(rc=%d, errno=%d strerror=%s)\n", rc, errno, strerror(errno));
exit (1);
}
}
}
}
#ifdef TIMING
//compute the timein us(10E-6)
end_t=__mftb();
printf("\n\nTotal time used by SPE: %f us\n", (float)((end_t-start_t)/79.8));
#endif
printf("\n\nPrinting matrix in 3 seconds..\n");
sleep(3);
printMatrix();
printf("\n\nTotal time used by SPE: %f us\n\n", (float)((end_t-start_t)/79.8));
}