int main()
{
xps_timer_t timer;
int time_create, time_start, time_stop;
// *************************************************************************************
extern unsigned char intermediate[];
extern unsigned int dct_handle_offset;
unsigned int dct_handle = (dct_handle_offset) + (unsigned int)(&intermediate);
extern unsigned int idct_handle_offset;
unsigned int idct_handle = (idct_handle_offset) + (unsigned int)(&intermediate);
// *************************************************************************************
// Thread attribute structures
Huint sta[NUM_THREADS];
void* retval[NUM_THREADS];
hthread_t tid[NUM_THREADS];
hthread_attr_t attr[NUM_THREADS];
targ_t targ[NUM_THREADS];
int my_dct_matrix[BLOCK_SIZE][BLOCK_SIZE] = {
{23170, 23170, 23170, 23170, 23170, 23170, 23170, 23170 },
{32138, 27246, 18205, 6393, -6393, -18205, -27246, -32138 },
{30274, 12540, -12540, -30274, -30274, -12540, 12540, 30274 },
{27246, -6393, -32138, -18205, 18205, 32138, 6393, -27246 },
{23170, -23170, -23170, 23170, 23170, -23170, -23170, 23170 },
{18205, -32138, 6393, 27246, -27246, -6393, 32138, -18205 },
{12540, -30274, 30274, -12540, -12540, 30274, -30274, 12540 },
{6393 , -18205, 27246, -32138, 32138, -27246, 18205, -6393 }
};
int my_dct_matrix_trans[BLOCK_SIZE][BLOCK_SIZE] = {
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0}
};
int my_temp[BLOCK_SIZE][BLOCK_SIZE];
int my_input[BLOCK_SIZE][BLOCK_SIZE];
int my_intermediate_dct[BLOCK_SIZE][BLOCK_SIZE];
int my_intermediate_idct[BLOCK_SIZE][BLOCK_SIZE];
int my_output[BLOCK_SIZE][BLOCK_SIZE];
int my_idct_output[BLOCK_SIZE][BLOCK_SIZE];
int my_scale_factor = 16;
// Create timer
xps_timer_create(&timer, (int*)0x20400000);
// Start timer
xps_timer_start(&timer);
int i,j;
#ifdef USE_MB_THREAD
printf("Using heterogeneous MB threads\r\n");
printf("\tCode start address = 0x%08x\n", (unsigned int)&intermediate);
#else
printf("Using native PPC threads\r\n");
#endif
for (j = 0; j < NUM_THREADS; j++)
{
// Initialize the attributes for the threads
hthread_attr_init( &attr[j] );
hthread_attr_sethardware( &attr[j], (void*)base_array[j] );
}
// Calculate transpose of dct_matrix
initialize_dct_matrix(my_dct_matrix, my_dct_matrix_trans);
// Initialize input
for (i = 0; i < BLOCK_SIZE; i++)
{
for (j = 0; j < BLOCK_SIZE; j++)
{
my_input[i][j] = i+j;
}
}
// Fill in thread arguments for an initial DCT run
targ[0].scale_factor = my_scale_factor;
targ[0].input = my_input;
targ[0].intermediate = my_intermediate_dct;
targ[0].output = my_temp;
targ[0].coeff_matrix = my_dct_matrix;
targ[0].coeff_matrix_trans = my_dct_matrix_trans;
dct_thread(&targ[0]);
printf("\r\nOriginal Input:\r\n");
print_matrix(my_input);
printf("\r\nOriginal DCT:\r\n");
print_matrix(my_temp);
printf("**************************************************\r\n");
// Fill in thread arguments
targ[0].scale_factor = my_scale_factor;
targ[0].input = my_input;
targ[0].intermediate = my_intermediate_dct;
targ[0].output = my_output;
targ[0].coeff_matrix = my_dct_matrix;
targ[0].coeff_matrix_trans = my_dct_matrix_trans;
targ[1].scale_factor = my_scale_factor;
targ[1].input = my_temp;
targ[1].intermediate = my_intermediate_idct;
targ[1].output = my_idct_output;
targ[1].coeff_matrix = my_dct_matrix;
targ[1].coeff_matrix_trans = my_dct_matrix_trans;
// Start timing thread create
time_create = xps_timer_read_counter(&timer);
// Peform DCT
//dct_thread(&targ);
// Use a thread
#ifdef USE_MB_THREAD
sta[0] = hthread_create( &tid[0], &attr[0], (void*)dct_handle, (void*)&targ[0] );
#else
sta[0] = hthread_create( &tid[0], NULL, dct_thread, (void*)&targ[0]);
#endif
// Peform IDCT
//idct_thread(&targ);
// Use a thread
#ifdef USE_MB_THREAD
sta[1] = hthread_create( &tid[1], &attr[1], (void*)idct_handle, (void*)&targ[1] );
#else
sta[1] = hthread_create( &tid[1], NULL, idct_thread, (void*)&targ[1]);
#endif
// Allow created threads to begin running and start timer
time_start = xps_timer_read_counter(&timer);
// Wait for threads to complete
hthread_join(tid[0],&retval[0]);
hthread_join(tid[1],&retval[1]);
// Stop timer
time_stop = xps_timer_read_counter(&timer);
printf("\r\nDCT (retval = 0x%08x):\r\n",(unsigned int)retval[0]);
print_matrix(my_output);
printf("\r\nIDCT (retval = 0x%08x):\r\n",(unsigned int)retval[1]);
print_matrix(my_idct_output);
printf("*********************************\n");
printf("Create time = %u\n",time_create);
printf("Start time = %u\n",time_start);
printf("Stop time = %u\n",time_stop);
printf("*********************************\n");
printf("Creation time (|Start - Create|) = %u\n",time_start - time_create);
printf("Elapsed time (|Stop - Start|) = %u\n",time_stop - time_start);
printf("Total time (|Create - Stop|) = %u\n",time_stop - time_create);
return 0;
}
int main( int argc, char *argv[] )
{
if (NUM_THREADS > 2){
printf("CANNOT USE MORE THAN (2) HETEROGENEOUS THREADS!!!!\r\n");
return (-1);
}
// Timer variables
xps_timer_t timer;
int time_create, time_start, time_stop;
// Mutex
hthread_mutex_t * mutex = (hthread_mutex_t*)malloc( sizeof(hthread_mutex_t) );
hthread_mutex_init( mutex, NULL );
// Thread attribute structures
Huint sta[NUM_THREADS];
void* retval[NUM_THREADS];
hthread_t tid[NUM_THREADS];
hthread_attr_t attr[NUM_THREADS];
targ_t thread_arg[NUM_THREADS];
// Setup Cache
XCache_EnableICache(0xc0000801);
// Create timer
xps_timer_create(&timer, (int*)0x20400000);
// Start timer
xps_timer_start(&timer);
// *************************************************************************************
extern unsigned char intermediate[];
extern unsigned int blink_handle_offset;
unsigned int blink_handle = (blink_handle_offset) + (unsigned int)(&intermediate);
// *************************************************************************************
int j = 0;
printf("Code start address = 0x%08x\n", (unsigned int)&intermediate);
for (j = 0; j < NUM_THREADS; j++)
{
// Initialize the attributes for the threads
hthread_attr_init( &attr[j] );
hthread_attr_sethardware( &attr[j], (void*)base_array[j] );
}
int num_ops = 0;
for( num_ops = 0; num_ops < 1; num_ops = num_ops + 1)
{
printf("******* Round %d ********\n",num_ops);
#ifdef USE_MB_THREAD
printf("**** MB-based Threads ****\n");
#else
printf("**** PPC-based Threads ****\n");
#endif
// Initialize thread arguments
for ( j= 0; j < NUM_THREADS; j++)
{
thread_arg[j].arg = 1 << ((j + 1));
thread_arg[j].mutex = mutex;
}
time_create = xps_timer_read_counter(&timer);
// Create threads
for (j = 0; j < NUM_THREADS; j++)
{
// Create the blink thread
#ifdef USE_MB_THREAD
// Create MB Thread
sta[j] = hthread_create( &tid[j], &attr[j], (void*)blink_handle, (void*)(&thread_arg[j]) );
//printf( "Started MB Thread (TID = %d) \n", tid[j]);
#else
// Create SW Thread
sta[j] = hthread_create( &tid[j], NULL, blink_thread, (void*)(&thread_arg[j]) );
//printf( "Started SW Thread (TID = %d) \n", tid[j]);
#endif
}
// Allow created threads to begin running and start timer
time_start = xps_timer_read_counter(&timer);
printf("Waiting for threads...\n");
// Join on all threads
for (j = 0; j < NUM_THREADS; j++)
{
hthread_join( tid[j], &retval[j] );
}
// Grab stop time
time_stop = xps_timer_read_counter(&timer);
// Print out status
for (j = 0; j < NUM_THREADS; j++)
{
printf("TID[%d] = 0x%08x, status = 0x%08x, retval = 0x%08x\n",j,tid[j],sta[j],(unsigned int)retval[j]);
}
printf("*********************************\n");
printf("Create time = %u\n",time_create);
printf("Start time = %u\n",time_start);
printf("Stop time = %u\n",time_stop);
printf("*********************************\n");
printf("Creation time (|Start - Create|) = %u\n",time_start - time_create);
printf("Elapsed time (|Stop - Start|) = %u\n",time_stop - time_start);
}
hthread_mutex_destroy( mutex );
free( mutex );
// Clean up the attribute structures
for (j = 0; j < NUM_THREADS; j++)
{
hthread_attr_destroy( &attr[j] );
}
printf ("-- Complete --\n");
// Return from main
return 0;
}
//int main( int argc, char *argv[] )
int run_tests()
{
if (NUM_THREADS > NUM_CPUS){
printf("CANNOT USE MORE THAN (%d) HETEROGENEOUS THREADS!!!!\r\n",NUM_CPUS);
return (-1);
}
// Timer variables
xps_timer_t timer;
int time_create, time_start, time_stop;
// Thread attribute structures
Huint sta[NUM_THREADS];
void* retval[NUM_THREADS];
hthread_t tid[NUM_THREADS];
hthread_attr_t attr[NUM_THREADS];
targ_t thread_arg[NUM_THREADS];
// Setup Cache
XCache_DisableDCache();
XCache_EnableICache(0xc0000801);
// Create timer
xps_timer_create(&timer, (int*)0x20400000);
// Start timer
xps_timer_start(&timer);
// *************************************************************************************
extern unsigned char intermediate[];
extern unsigned int sort_handle_offset;
unsigned int sort_handle = (sort_handle_offset) + (unsigned int)(&intermediate);
// *************************************************************************************
float local_data[ARR_LENGTH];
int j = 0;
printf("Code start address = 0x%08x\n", (unsigned int)&intermediate);
for (j = 0; j < NUM_THREADS; j++)
{
// Initialize the attributes for the threads
hthread_attr_init( &attr[j] );
hthread_attr_sethardware( &attr[j], (void*)base_array[j] );
}
for (j = 0; j < ARR_LENGTH; j++)
{
local_data[j] = (ARR_LENGTH - j)*1.00;
}
#ifdef MY_DEBUG
show_array(&local_data[0], ARR_LENGTH);
#endif
int num_ops = 0;
for( num_ops = 0; num_ops < 2; num_ops = num_ops + 1)
{
printf("******* Round %d ********\n",num_ops);
#ifdef USE_MB_THREAD
printf("**** MB-based Threads ****\n");
#else
printf("**** PPC-based Threads ****\n");
#endif
for (j = 0; j < ARR_LENGTH; j++)
{
local_data[j] = ARR_LENGTH - j;
}
// Initialize thread arguments
for ( j= 0; j < NUM_THREADS; j++)
{
thread_arg[j].data = &local_data[j*(ARR_LENGTH/NUM_THREADS)];
thread_arg[j].length = ARR_LENGTH/NUM_THREADS;
}
time_create = xps_timer_read_counter(&timer);
// Create threads
for (j = 0; j < NUM_THREADS; j++)
{
// Create the sort thread
#ifdef USE_MB_THREAD
// Create MB Thread
sta[j] = hthread_create( &tid[j], &attr[j], (void*)sort_handle, (void*)(&thread_arg[j]) );
//printf( "Started MB Thread (TID = %d) \n", tid[j]);
#else
// Create SW Thread
sta[j] = hthread_create( &tid[j], NULL, bubblesort_thread, (void*)(&thread_arg[j]) );
//printf( "Started SW Thread (TID = %d) \n", tid[j]);
#endif
}
// Allow created threads to begin running and start timer
time_start = xps_timer_read_counter(&timer);
// Join on all threads
for (j = 0; j < NUM_THREADS; j++)
{
hthread_join( tid[j], &retval[j] );
}
// Grab stop time
time_stop = xps_timer_read_counter(&timer);
// Print out status
for (j = 0; j < NUM_THREADS; j++)
{
printf("TID[%d] = 0x%08x, status = 0x%08x, retval = 0x%08x\n",j,tid[j],sta[j],(unsigned int)retval[j]);
}
printf("*********************************\n");
printf("Create time = %u\n",time_create);
printf("Start time = %u\n",time_start);
printf("Stop time = %u\n",time_stop);
printf("*********************************\n");
printf("Creation time (|Start - Create|) = %u\n",time_start - time_create);
printf("Elapsed time (|Stop - Start|) = %u\n",time_stop - time_start);
printf("Total time (|Create - Stop|) = %u\n",time_stop - time_create);
}
#ifdef MY_DEBUG
show_array(&local_data[0], ARR_LENGTH);
check_array(&local_data[0], ARR_LENGTH); // Note there will be errors as the data set is not merged
#endif
// Clean up the attribute structures
for (j = 0; j < NUM_THREADS; j++)
{
hthread_attr_destroy( &attr[j] );
}
printf ("-- Complete --\n");
// Return from main
return 0;
}
int run_tests()
{
if (NUM_THREADS > NUM_CPUS){
printf("CANNOT USE MORE THAN (%d) HETEROGENEOUS THREADS!!!!\r\n", NUM_CPUS);
return (-1);
}
// Timer variables
xps_timer_t timer;
int time_create, time_start, time_stop;
// Thread attribute structures
Huint sta[NUM_THREADS];
void* retval[NUM_THREADS];
hthread_t tid[NUM_THREADS];
hthread_attr_t attr[NUM_THREADS];
targ_t thread_arg[NUM_THREADS];
// Setup Cache
XCache_DisableDCache();
XCache_EnableICache(0xc0000801);
// Create timer
xps_timer_create(&timer, (int*)0x20400000);
// Start timer
xps_timer_start(&timer);
// *************************************************************************************
extern unsigned char intermediate[];
extern unsigned int distance_handle_offset;
unsigned int distance_handle = (distance_handle_offset) + (unsigned int)(&intermediate);
printf("Code start address = 0x%08x\n", (unsigned int)&intermediate);
// *************************************************************************************
int vals_x0[ARR_LENGTH];
int vals_x1[ARR_LENGTH];
int vals_y0[ARR_LENGTH];
int vals_y1[ARR_LENGTH];
int vals_ds[ARR_LENGTH];
int j = 0;
for (j = 0; j < NUM_THREADS; j++)
{
// Initialize the attributes for the threads
hthread_attr_init( &attr[j] );
hthread_attr_sethardware( &attr[j], (void*)base_array[j] );
}
for (j = 0; j < ARR_LENGTH; j++)
{
vals_x0[j] = ARR_LENGTH - j;
vals_y0[j] = ARR_LENGTH - j;
vals_x1[j] = ARR_LENGTH - j + 1;
vals_y1[j] = ARR_LENGTH - j;
}
#ifdef MY_DEBUG
#endif
int num_ops = 0;
for( num_ops = 0; num_ops < 2; num_ops = num_ops + 1)
{
printf("******* Round %d ********\n",num_ops);
#ifdef USE_MB_THREAD
printf("**** MB-based Threads ****\n");
#else
printf("**** PPC-based Threads ****\n");
#endif
for (j = 0; j < ARR_LENGTH; j++)
{
vals_x0[j] = ARR_LENGTH - j;
}
// Initialize thread arguments
int num_items = ARR_LENGTH/NUM_THREADS;
int extra_items = ARR_LENGTH - (num_items*NUM_THREADS);
for ( j= 0; j < NUM_THREADS; j++)
{
thread_arg[j].x0s = &vals_x0[j*(num_items)];
thread_arg[j].y0s = &vals_y0[j*(num_items)];
thread_arg[j].x1s = &vals_x1[j*(num_items)];
thread_arg[j].y1s = &vals_y1[j*(num_items)];
thread_arg[j].distances = &vals_ds[j*(num_items)];
thread_arg[j].length = num_items;
}
// Add in extra items for the last thread if needed
thread_arg[j-1].length += extra_items;
time_create = xps_timer_read_counter(&timer);
// Create threads
for (j = 0; j < NUM_THREADS; j++)
{
// Create the distance thread
#ifdef USE_MB_THREAD
// Create MB Thread
sta[j] = hthread_create( &tid[j], &attr[j], (void*)distance_handle, (void*)(&thread_arg[j]) );
//printf( "Started MB Thread (TID = %d) \n", tid[j]);
#else
// Create SW Thread
sta[j] = hthread_create( &tid[j], NULL, distance_thread, (void*)(&thread_arg[j]) );
//printf( "Started SW Thread (TID = %d) \n", tid[j]);
#endif
}
// Allow created threads to begin running and start timer
time_start = xps_timer_read_counter(&timer);
// Join on all threads
for (j = 0; j < NUM_THREADS; j++)
{
hthread_join( tid[j], &retval[j] );
}
// Grab stop time
time_stop = xps_timer_read_counter(&timer);
// Print out status
for (j = 0; j < NUM_THREADS; j++)
{
printf("TID[%d] = 0x%08x, status = 0x%08x, retval = 0x%08x\n",j,tid[j],sta[j],(unsigned int)retval[j]);
}
printf("*********************************\n");
printf("Create time = %u\n",time_create);
printf("Start time = %u\n",time_start);
printf("Stop time = %u\n",time_stop);
printf("*********************************\n");
printf("Creation time (|Start - Create|) = %u\n",time_start - time_create);
printf("Elapsed time (|Stop - Start|) = %u\n",time_stop - time_start);
printf("Total time (|Create - Stop|) = %u\n",time_stop - time_create);
}
#ifdef MY_DEBUG
for (j = 0; j < ARR_LENGTH; j++)
{
printf("D(%d) = %d\n",j,vals_ds[j]);
}
#endif
// Clean up the attribute structures
for (j = 0; j < NUM_THREADS; j++)
{
hthread_attr_destroy( &attr[j] );
}
printf ("-- Complete --\n");
// Return from main
return 0;
}
int run_tests()
{
// Timer variables
xps_timer_t timer;
int time_create, time_start, time_unlock, time_stop;
// Mutex
hthread_mutex_t * mutex = (hthread_mutex_t*)malloc( sizeof(hthread_mutex_t) );
hthread_mutex_init( mutex, NULL );
float min;
// Thread attribute structures
Huint sta[NUM_THREADS];
void* retval[NUM_THREADS];
hthread_t tid[NUM_THREADS];
hthread_attr_t attr[NUM_THREADS];
targ_t thread_arg[NUM_THREADS];
// Setup Cache
XCache_DisableDCache();
XCache_EnableICache(0xc0000801);
// Create timer
xps_timer_create(&timer, (int*)0x20400000);
// Start timer
xps_timer_start(&timer);
// *************************************************************************************
extern unsigned char intermediate[];
extern unsigned int min_handle_offset;
unsigned int min_handle = (min_handle_offset) + (unsigned int)(&intermediate);
// *************************************************************************************
printf("Code start address = 0x%08x\n", (unsigned int)&intermediate);
int i = 0;
float main_array[ARRAY_LENGTH];
printf("Addr of array = 0x%08x\n",(unsigned int)&main_array[0]);
for (i = 0; i < ARRAY_LENGTH; i++)
{
main_array[i] = (i+2)*3.14f;
}
int num_items = ARRAY_LENGTH/NUM_THREADS;
int extra_items = ARRAY_LENGTH - (num_items*NUM_THREADS);
float * start_addr = &main_array[0];
for (i = 0; i < NUM_THREADS; i++)
{
// Initialize the attributes for the hardware threads
hthread_attr_init( &attr[i] );
hthread_attr_sethardware( &attr[i], (void*)base_array[i] );
// Initialize thread arguments
thread_arg[i].num_items = num_items;
thread_arg[i].data_ptr = start_addr;
thread_arg[i].min_mutex = mutex;
thread_arg[i].min = &min;
start_addr+=num_items;
}
// Add in extra items for the last thread if needed
thread_arg[i-1].num_items += extra_items;
int num_ops = 0;
for( num_ops = 0; num_ops < 2; num_ops = num_ops + 1)
{
printf("******* Round %d ********\n",num_ops);
#ifdef USE_MB_THREAD
printf("**** MB-based Threads ****\n");
#else
printf("**** PPC-based Threads ****\n");
#endif
min = 9999999;
// Lock mutex before hand so that timing will not include thread creation time
hthread_mutex_lock(mutex);
// Start timing thread create
time_create = xps_timer_read_counter(&timer);
for (i = 0; i < NUM_THREADS; i++)
{
// Create the worker threads
#ifdef USE_MB_THREAD
// Create MB Thread
sta[i] = hthread_create( &tid[i], &attr[i], (void*)(min_handle), (void*)(&thread_arg[i]) );
#else
// Create SW Thread
sta[i] = hthread_create( &tid[i], NULL, min_thread, (void*)(&thread_arg[i]) );
#endif
}
// Allow created threads to begin running and start timer
time_start = xps_timer_read_counter(&timer);
hthread_mutex_unlock(mutex);
time_unlock = xps_timer_read_counter(&timer);
// Wait for the threads to exit
//printf( "Waiting for thread(s) to complete... \n" );
for (i = 0; i < NUM_THREADS; i++)
{
hthread_join( tid[i], &retval[i] );
}
time_stop = xps_timer_read_counter(&timer);
// Display results
printf("Min = %f\n",min);
for (i = 0; i < NUM_THREADS; i++)
{
printf("TID = 0x%08x, status = 0x%08x, retval = 0x%08x\n",tid[i],sta[i],(Huint)retval[i]);
}
printf("*********************************\n");
printf("Create time = %u\n",time_create);
printf("Start time = %u\n",time_start);
printf("Unlock time = %u\n",time_unlock);
printf("Stop time = %u\n",time_stop);
printf("*********************************\n");
printf("Creation time (|Start - Create|) = %u\n",time_start - time_create);
printf("Unlock time (|Unlock - Start|) = %u\n",time_unlock - time_start);
printf("Elapsed time (|Stop - Start|) = %u\n",time_stop - time_start);
}
hthread_mutex_destroy( mutex );
free( mutex );
// Clean up the attribute structures
for (i = 0; i < NUM_THREADS; i++)
{
hthread_attr_destroy( &attr[i] );
}
printf ("-- Complete --\n");
// Return from main
return 0;
}
