/**
*
* This function is used to override the driver's default sleep functionality.
* For MicroBlaze systems, the XDpRxSs_WaitUs driver function's default behavior
* is to use the MB_Sleep function from microblaze_sleep.h, which is implemented
* in software and only has millisecond accuracy. For this reason, using a
* hardware timer is preferable. For ARM/Zynq SoC systems, the SoC's timer is
* used - XDpRxSs_WaitUs will ignore this custom timer handler.
*
* @param InstancePtr is a pointer to the XDpRxSs core instance.
* @param MicroSeconds is the number of microseconds to delay/sleep for.
*
* @return None.
*
* @note Use the XDpRxSs_SetUserTimerHandler driver function to set this
* function as the handler for when the XDpRxSs_WaitUs driver
* function is called.
*
******************************************************************************/
void DpRxSs_CustomWaitUs(void *InstancePtr, u32 MicroSeconds)
{
XDpRxSs *DpRxSsPtr = (XDpRxSs *)InstancePtr;
u32 TimerVal;
XTmrCtr_Start(DpRxSsPtr->DpPtr->UserTimerPtr, 0);
/* Wait specified number of useconds. */
do {
TimerVal = XTmrCtr_GetValue(DpRxSsPtr->DpPtr->UserTimerPtr, 0);
} while (TimerVal < (MicroSeconds *
(DpRxSsPtr->DpPtr->Config.SAxiClkHz / 1000000)));
XTmrCtr_Stop(DpRxSsPtr->DpPtr->UserTimerPtr, 0);
}
int main (void)
{
float u[n*n], v[n*n], u0[n*n], v0[n*n];
float x, y, x0, y_0, f, r, U[2], V[2], s, t;
int i, j, i0, j_0, i1, j_1,m;
int k;
float visc=1.25;
float dt=2.36;
int start,end;
xil_printf("----------------------- START ------------------------ ! \n\r");
int MASK_LENGTH= 10;//(int) ceil((log(BUFFER_SIZE))/log(2));
int MASK = ((1 << MASK_LENGTH)-1);
//-- Flag Initializations
for (i=0; i<BUFFER_SIZE; i++)
{
flag1[i]=0;
flag2[i]=0;
flag3[i]=0;
}
XTmrCtr timer;
XTmrCtr_Initialize(&timer, XPAR_XPS_TIMER_0_DEVICE_ID);
XTmrCtr_Reset(&timer, TIMER_COUNTER_0);
start = XTmrCtr_GetValue(&timer, TIMER_COUNTER_0);
// Start timer
XTmrCtr_Start(&timer, TIMER_COUNTER_0);
//------- LOOP FOR IMAGES -------
for (k=0; k < Number_of_Images; k++) // loop for Images
{
consumer_index = ci; // for nested loop can be i*n+j
while (consumer_index<SIZE_floid)
{
if (read_from_fifo ==1)
{
getfsl(fifo_index,0);
if (fifo_index!=999999999)
{
getfsl(fifo_data,0);
getfsl(fifo_data2,0);
xil_printf("READ %d FROM FIFO \t",fifo_index);
if (consumer_index == fifo_index && fifo_index!=999999999)
{
// consume the data directly
consumer_data = fifo_data;
consumer_data2 = fifo_data2;
xil_printf("»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»» %d DIRECTLY CONSUMED !\n\r",consumer_index);
consumer_index=consumer_index+1;//ci;
consumer_done=1;
read_from_fifo=1;
//-- The rest of consumer's computing stage here : output: consumer_index
u0[i+(n+2)*j] = consumer_data; // u[i+n*j]
v0[i+(n+2)*j] = consumer_data2; // v[i+n*j]
y = j<=n/2 ? j : j-n;
r = x*x+y*y;
if ( r==0.0 ) continue;
f = exp(-r*dt*visc);
U[0] = u0[i +(n+2)*j];
V[0] = v0[i +(n+2)*j];
U[1] = u0[i+1+(n+2)*j];
V[1] = v0[i+1+(n+2)*j];
u0[i +(n+2)*j] = f*( (1-x*x/r)*U[0] -x*y/r *V[0] );
u0[i+1+(n+2)*j] = f*( (1-x*x/r)*U[1] -x*y/r *V[1] );
v0[i+ (n+2)*j] = f*( -y*x/r *U[0] + (1-y*y/r)*V[0] );
v0[i+1+(n+2)*j] = f*( -y*x/r *U[1] + (1-y*y/r)*V[1] );
}
}
else if (fifo_index==999999999)
{
xil_printf("==PRODUCER STOPPED AND consumer_index IS %d!\n\r",consumer_index);
read_from_fifo=0;
//consumer_index++;
}
}
// compute the hash
consumer_hash_index = consumer_index & MASK; // here index is not refreshing
fifo_hash_index = fifo_index & MASK;
// Load from memory
if (flag1[consumer_hash_index]==1 && index1[consumer_hash_index]==consumer_index)
{
flag1[consumer_hash_index]=0;
consumer_data= data1[consumer_hash_index];
xil_printf("========== LOAD consumer_index %d from T1 ================!\n\r",consumer_index);
consumer_index++;
consumer_done=1;
//-- The rest of consumer's computing stage here : output: consumer_index
u0[i+(n+2)*j] = consumer_data; // u[i+n*j]
v0[i+(n+2)*j] = consumer_data2; // v[i+n*j]
y = j<=n/2 ? j : j-n;
r = x*x+y*y;
if ( r==0.0 ) continue;
f = exp(-r*dt*visc);
U[0] = u0[i +(n+2)*j];
V[0] = v0[i +(n+2)*j];
U[1] = u0[i+1+(n+2)*j];
V[1] = v0[i+1+(n+2)*j];
u0[i +(n+2)*j] = f*( (1-x*x/r)*U[0] -x*y/r *V[0] );
u0[i+1+(n+2)*j] = f*( (1-x*x/r)*U[1] -x*y/r *V[1] );
v0[i+ (n+2)*j] = f*( -y*x/r *U[0] + (1-y*y/r)*V[0] );
v0[i+1+(n+2)*j] = f*( -y*x/r *U[1] + (1-y*y/r)*V[1] );
if (fifo_index!=999999999)
read_from_fifo=1;
}
else if (flag2[consumer_hash_index]==1 && index2[consumer_hash_index]==consumer_index)
{
flag2[consumer_hash_index]=0;
consumer_data= data2[consumer_hash_index];
consumer_index++;
read_from_fifo=1;
consumer_done=1;
xil_printf("=================== LOAD consumer_index %d from T2 ================!\n\r",consumer_index);
//-- The rest of consumer's computing stage here : output: consumer_index
u0[i+(n+2)*j] = consumer_data; // u[i+n*j]
v0[i+(n+2)*j] = consumer_data2; // v[i+n*j]
y = j<=n/2 ? j : j-n;
r = x*x+y*y;
if ( r==0.0 ) continue;
f = exp(-r*dt*visc);
U[0] = u0[i +(n+2)*j];
V[0] = v0[i +(n+2)*j];
U[1] = u0[i+1+(n+2)*j];
V[1] = v0[i+1+(n+2)*j];
u0[i +(n+2)*j] = f*( (1-x*x/r)*U[0] -x*y/r *V[0] );
u0[i+1+(n+2)*j] = f*( (1-x*x/r)*U[1] -x*y/r *V[1] );
v0[i+ (n+2)*j] = f*( -y*x/r *U[0] + (1-y*y/r)*V[0] );
v0[i+1+(n+2)*j] = f*( -y*x/r *U[1] + (1-y*y/r)*V[1] );
}
else if (flag3[consumer_hash_index]==1 && index3[consumer_hash_index]==consumer_index)
{
flag3[consumer_hash_index]=0;
consumer_data= data3[consumer_hash_index];
consumer_index++;
read_from_fifo=1;
consumer_done=1;
consumer_done=1;
xil_printf("=================== LOAD consumer_index %d from T3 ================!\n\r",consumer_index);
//-- The rest of consumer's computing stage here : output: consumer_index
u0[i+(n+2)*j] = consumer_data; // u[i+n*j]
v0[i+(n+2)*j] = consumer_data2; // v[i+n*j]
y = j<=n/2 ? j : j-n;
r = x*x+y*y;
if ( r==0.0 ) continue;
f = exp(-r*dt*visc);
U[0] = u0[i +(n+2)*j];
V[0] = v0[i +(n+2)*j];
U[1] = u0[i+1+(n+2)*j];
V[1] = v0[i+1+(n+2)*j];
u0[i +(n+2)*j] = f*( (1-x*x/r)*U[0] -x*y/r *V[0] );
u0[i+1+(n+2)*j] = f*( (1-x*x/r)*U[1] -x*y/r *V[1] );
v0[i+ (n+2)*j] = f*( -y*x/r *U[0] + (1-y*y/r)*V[0] );
v0[i+1+(n+2)*j] = f*( -y*x/r *U[1] + (1-y*y/r)*V[1] );
}
// Store in memory
else if (read_from_fifo ==1 && consumer_done==0)
{
if (flag1[fifo_hash_index]==0 && fifo_index!=999999999)
{
data1[fifo_hash_index] =fifo_data;
index1[fifo_hash_index] =fifo_index;
flag1[fifo_hash_index]=1;
read_from_fifo=1;
xil_printf("»»»»»»»»»»»»»»»»»%d STORED in T1 \n\r",fifo_index);
}
else if (flag2[fifo_hash_index]==0 && fifo_index!=999999999)
{
data2[fifo_hash_index] =fifo_data;
index2[fifo_hash_index] =fifo_index;
flag2[fifo_hash_index]=1;
read_from_fifo=1;
xil_printf("»»»»»»»»»»»»»»»»»%d STORED in T2 \n\r",fifo_index);
}
else if (flag3[fifo_hash_index]==0 && fifo_index!=999999999)
{
data3[fifo_hash_index] =fifo_data;
index3[fifo_hash_index] =fifo_index;
flag3[fifo_hash_index]=1;
read_from_fifo=1;
xil_printf("»»»»»»»»»»»»»»»»»%d STORED in T3 \n\r",fifo_index);
}
else
read_from_fifo=0;
} // Store in memory
else
{
xil_printf("ERROR ! Index %d cannot be found !!!!\n\r",consumer_index);
consumer_done=0;
consumer_index++;
}
}
}
xil_printf ("consumer_index =%d \t ci=%d \n\r",consumer_index,ci);
end = XTmrCtr_GetValue(&timer, TIMER_COUNTER_0);
XTmrCtr_Stop(&timer, TIMER_COUNTER_0);
xil_printf("Timer Start value = %d Timer end value = %d \r\n", start, end-start);
return 0;
}
u32 getTimerValue(u8 TmrCtrNumber) {
return XTmrCtr_GetValue(&timer, TmrCtrNumber);
}
void * foo_thread(void * arg)
{
data * package = (data *) arg;
Hint * dataA = package->dataA;
Hint * dataB = package->dataB;
Hint * dataC = package->dataC;
Hint * brama = (Hint *) BRAMA;
Hint * bramb = (Hint *) BRAMB;
Hint * bramc = (Hint *) BRAMC;
int e,Status;
int time = package->time;
XTmrCtr * mytimer =package->timer;
XAxiCdma mydma;
#ifndef HETERO_COMPILATION
Status= XAxiCdma_Initialize(&mydma, XPAR_PERIPHERALS_CENTRAL_DMA_DEVICE_ID);
#else
Status= XAxiCdma_Initialize(&mydma,XPAR_GROUP_0_SLAVE_0_LOCAL_DMA_DEVICE_ID);
#endif
if (Status != XST_SUCCESS) {
putnum(0xdeadbeef);
return XST_FAILURE;
}
XAxiCdma_IntrDisable(&mydma, XAXICDMA_XR_IRQ_ALL_MASK);
XTmrCtr_Reset(mytimer,1);
//sending the data **********************************************
/*for (e = 0; e < package->vector_size; e++) {
brama[e] = dataA[e];
bramb[e] = dataB[e]; }
*/
Status = XAxiCdma_Transfer(&mydma, (u32) package->dataA , (u32) BRAMA, package->vector_size *4, NULL, NULL);
Status = XAxiCdma_Transfer(&mydma, (u32) package->dataB , (u32) BRAMB, package->vector_size *4, NULL, NULL);
//computation
// for (e = 0; e < package->vector_size; e++) bramc[e] = brama[e] + bramb[e];
/*
int cmd =1; //1=> means add, 2=> means subtraction
int start =0;
int end = package->vector_size ;
putfslx( cmd, 0, FSL_DEFAULT);
putfslx( end, 0, FSL_DEFAULT);
putfslx( start, 0, FSL_DEFAULT);
getfslx(cmd, 0, FSL_DEFAULT);
*/
putfslx((0 << 30) | (SIZE << 15) | 0, 0, FSL_DEFAULT);
getfslx(e, 0, FSL_DEFAULT);
//sending the data back to dram **********************************************
Status = XAxiCdma_Transfer(&mydma, (u32) BRAMC , (u32) package->dataC, package->vector_size *4, NULL, NULL);
// for (e = 0; e < package->vector_size; e++) dataC[e] = bramc[e];
package->time= XTmrCtr_GetValue(mytimer,1);
return (void *) 0;
}
int main(int argc, char *argv[])
{
xil_printf("\r\n-----Begin Program-----\r\n");
//===============================================================================
//Initilizing Data
//===============================================================================
data package;
package.vector_size = SIZE;
package.dataA = (int *) malloc(sizeof(int) * package.vector_size);
package.dataB = (int *) malloc(sizeof(int) * package.vector_size);
package.dataC = (int *) malloc(sizeof(int) * package.vector_size);
// xil_printf (" %08x , %08x , %08x , \r\n", package.dataA,package.dataB,package.dataC);
assert(package.dataA != NULL);
assert(package.dataB != NULL);
assert(package.dataC != NULL);
int e;
Hint * dataA = package.dataA;
Hint * dataB = package.dataB;
Hint * dataC = package.dataC;
// for (e = 0; e < package.vector_size; e++) xil_printf("%d %d %d\r\n", dataA[e], dataB[e], dataC[e]);
//===============================================================================
//Peripherals instantiation
//===============================================================================
XTmrCtr mytimer;
int Status= XTmrCtr_Initialize(&mytimer, XPAR_PERIPHERALS_AXI_TIMER_0_DEVICE_ID);
if (Status != XST_SUCCESS) {
xil_printf("Timer initialization failed \r\n");
return XST_FAILURE;
}
XTmrCtr_SetResetValue(&mytimer, 0, 0);
XTmrCtr_Start(&mytimer,0);
XTmrCtr_SetResetValue(&mytimer, 1, 0);
XTmrCtr_Start(&mytimer,1);
package.timer= &mytimer;
//===============================================================================
//Creating and joining on thread
//===============================================================================
hthread_t tid;
hthread_attr_t attr;
hthread_attr_init(&attr);
int i;
for ( i=0; i <NUM_AVAILABLE_HETERO_CPUS ; i++)
{
for (e = 0; e < package.vector_size; e++)
{
dataA[e] = rand() %1000;
dataB[e] = rand()%1000;
dataC[e] = 10;
}
XTmrCtr_Reset(&mytimer,0);
thread_create ( &tid, &attr , foo_thread_FUNC_ID , (void * )&package, i+2, 0);
int ret;
if( hthread_join(tid, (void *) &ret))
{
xil_printf("Error joining child thread\r\n");
while(1);
}
if (ret !=XST_SUCCESS){
xil_printf("Thread returned XST_FAILURE\r\n");
while(1);
}
int time= XTmrCtr_GetValue(&mytimer,0);
xil_printf("Total exe_time on Slave %d : %d us\r\n", i, (time)/100 );
xil_printf("Net exe_time of the body of the thread: %d us\r\n", (package.time)/100 );
//===============================================================================
//Check to see if the result were right?
//===============================================================================
for (e = 0; e < 5; e++) xil_printf("%d %d %d\r\n", dataA[e], dataB[e], dataC[e]);
for (e = 0; e < package.vector_size; e++)
{
if (dataC[e]!= dataA[e]+ dataB[e])
{
xil_printf("Error at location %d \r\n", e);
while (1);
}
}
}
xil_printf("FINISH\r\n");
return 0;
}
void Timer_Start(u8 id)
{
s_log[id].start_time = XTmrCtr_GetValue(&xtmrctr, 0);
}
void Timer(u8* String)
{
s_log[counter].time =XTmrCtr_GetValue(&xtmrctr, 0);
memcpy(s_log[counter++].msg, String, strlen((char*)String));
}
