int main(int argc, char** argv)
{
int step,burst;
int nparticle = 8192; /* MUST be a nice power of two for simplicity */
int nstep = 500;
int nburst = 20; /* MUST divide the value of nstep without remainder */
int nthread = 64; /* chosen for ATI Radeon HD 5870 */
float dt = 0.0001;
float eps = 0.0001;
cl_float4* pos1 = (cl_float4*)clmalloc(stdgpu,nparticle*sizeof(cl_float4),0);
cl_float4* pos2 = (cl_float4*)clmalloc(stdgpu,nparticle*sizeof(cl_float4),0);
cl_float4* vel = (cl_float4*)clmalloc(stdgpu,nparticle*sizeof(cl_float4),0);
nbody_init(nparticle,pos1,vel);
void* h = clopen(stdgpu,"nbody_kern.cl",CLLD_NOW);
cl_kernel krn = clsym(stdgpu,h,"nbody_kern",CLLD_NOW);
clndrange_t ndr = clndrange_init1d(0,nparticle,nthread);
clarg_set(stdgpu,krn,0,dt);
clarg_set(stdgpu,krn,1,eps);
clarg_set_global(stdgpu,krn,4,vel);
clarg_set_local(stdgpu,krn,5,nthread*sizeof(cl_float4));
clmsync(stdgpu,0,pos1,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync(stdgpu,0,vel,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
for(step=0; step<nstep; step+=nburst) {
for(burst=0; burst<nburst; burst+=2) {
clarg_set_global(stdgpu,krn,2,pos1);
clarg_set_global(stdgpu,krn,3,pos2);
clfork(stdgpu,0,krn,&ndr,CL_EVENT_NOWAIT);
clarg_set_global(stdgpu,krn,2,pos2);
clarg_set_global(stdgpu,krn,3,pos1);
clfork(stdgpu,0,krn,&ndr,CL_EVENT_NOWAIT);
}
clmsync(stdgpu,0,pos1,CL_MEM_HOST|CL_EVENT_NOWAIT);
clwait(stdgpu,0,CL_KERNEL_EVENT|CL_MEM_EVENT);
}
nbody_output(nparticle,pos1,vel);
clclose(stdgpu,h);
clfree(pos1);
clfree(pos2);
clfree(vel);
}
BasketGeometricOpenCLOption::~BasketGeometricOpenCLOption()
{
free(start_prices);
free(asset_volatilities);
free(correlations);
if(cl_start_prices) {
clfree(cl_start_prices);
}
if(cl_asset_volatilities) {
clfree(cl_asset_volatilities);
}
if(cl_correlations) {
clfree(cl_correlations);
}
}
int
uhidclose(dev_t dev, int flag, int mode, usb_proc_ptr p)
{
struct uhid_softc *sc;
USB_GET_SC(uhid, UHIDUNIT(dev), sc);
DPRINTF(("uhidclose: sc=%p\n", sc));
/* Disable interrupts. */
usbd_abort_pipe(sc->sc_intrpipe);
usbd_close_pipe(sc->sc_intrpipe);
sc->sc_intrpipe = 0;
ndflush(&sc->sc_q, sc->sc_q.c_cc);
clfree(&sc->sc_q);
free(sc->sc_ibuf, M_USBDEV);
free(sc->sc_obuf, M_USBDEV);
sc->sc_state &= ~UHID_OPEN;
sc->sc_async = 0;
return (0);
}
void BasketGeometricOpenCLOption::cleanup()
{
if(cl_start_prices) {
clfree(cl_start_prices);
}
if(cl_asset_volatilities) {
clfree(cl_asset_volatilities);
}
if(cl_correlations) {
clfree(cl_correlations);
}
cl_start_prices = NULL;
cl_asset_volatilities = NULL;
cl_correlations = NULL;
}
/*
* Tests memory allocation on the GPU by reading raster data in and,
* after transfering it to the device, read it back and creating an
* output raster that should be identical to the input one.
*
* int nrows .............. the number of rows in the DEM.
* int ncols .............. the number of columns in the DEM.
* int infd ............... file descriptor to read from the DEM.
* int outfd .............. file descriptor to write to an output raster.
* RASTER_MAP_TYPE rtype .. DEM element data type.
*
*/
void test_cl (const int nrows, const int ncols,
const int infd, const int outfd,
RASTER_MAP_TYPE rtype)
{
int row, col;
CONTEXT * ctx = init_context ( );
/* load raster DEM data */
FCELL *dem_device = load_raster (ctx, nrows, ncols, infd, rtype);
G_message ("Transfering raster data to the device ...");
clmsync (ctx, 0, dem_device, CL_MEM_DEVICE|CL_EVENT_WAIT);
G_message ("Transfering data from the device to the host ...");
clmsync (ctx, 0, dem_device, CL_MEM_HOST|CL_EVENT_WAIT);
/* save raster data */
save_raster (dem_device, nrows, ncols, outfd);
/* free allocated resources */
clfree (dem_device);
}
int
ugen_do_close(struct ugen_softc *sc, int endpt, int flag)
{
struct ugen_endpoint *sce;
int dir, i;
#ifdef DIAGNOSTIC
if (!sc->sc_is_open[endpt]) {
printf("ugenclose: not open\n");
return (EINVAL);
}
#endif
if (endpt == USB_CONTROL_ENDPOINT) {
DPRINTFN(5, ("ugenclose: close control\n"));
sc->sc_is_open[endpt] = 0;
return (0);
}
for (dir = OUT; dir <= IN; dir++) {
if (!(flag & (dir == OUT ? FWRITE : FREAD)))
continue;
sce = &sc->sc_endpoints[endpt][dir];
if (sce == NULL || sce->pipeh == NULL)
continue;
DPRINTFN(5, ("ugenclose: endpt=%d dir=%d sce=%p\n",
endpt, dir, sce));
usbd_close_pipe(sce->pipeh);
sce->pipeh = NULL;
switch (sce->edesc->bmAttributes & UE_XFERTYPE) {
case UE_INTERRUPT:
ndflush(&sce->q, sce->q.c_cc);
clfree(&sce->q);
break;
case UE_ISOCHRONOUS:
for (i = 0; i < UGEN_NISOREQS; ++i)
usbd_free_xfer(sce->isoreqs[i].xfer);
default:
break;
}
if (sce->ibuf != NULL) {
free(sce->ibuf, M_USBDEV, 0);
sce->ibuf = NULL;
}
}
sc->sc_is_open[endpt] = 0;
return (0);
}
void BasketGeometricOpenCLOption::setup_inputs()
{
if(cl_start_prices) {
clfree(cl_start_prices);
}
if(cl_asset_volatilities) {
clfree(cl_asset_volatilities);
}
if(cl_correlations) {
clfree(cl_correlations);
}
cl_start_prices = opencl_memcpy<cl_float, float>(context, number_of_assets, start_prices);
cl_asset_volatilities = opencl_memcpy<cl_float, float>(context, number_of_assets, asset_volatilities);
cl_correlations = opencl_memcpy<cl_float, float>(context, number_of_assets*number_of_assets, correlations);
assert(cl_start_prices != NULL);
assert(cl_asset_volatilities != NULL);
assert(cl_correlations != NULL);
clmsync(context, device_number, cl_start_prices, CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync(context, device_number, cl_asset_volatilities, CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync(context, device_number, cl_correlations, CL_MEM_DEVICE|CL_EVENT_NOWAIT);
}
int
uhidclose(dev_t dev, int flag, int mode, struct proc *p)
{
struct uhid_softc *sc;
sc = uhid_cd.cd_devs[UHIDUNIT(dev)];
DPRINTF(("uhidclose: sc=%p\n", sc));
clfree(&sc->sc_q);
free(sc->sc_obuf, M_USBDEV);
sc->sc_async = NULL;
uhidev_close(&sc->sc_hdev);
return (0);
}
void cexit() {
printf(ANSI_COLOR_BLUE"Saliendo!" ANSI_COLOR_RESET "\n");
clfree();
handOff(0);
exit(0);
}
int main (int argc, char * argv[])
{
char kernFile[] = "../src/nD.cl"; // the default location of the cl file - can be over ridden by the first command line arguement
char debugFile[] = "./debug"; // the debug file location
char msg[255]; // space for the debug message where a program variable is to be written to the debug file
char * clFile; // the cl file string actually used (either argv or kernFile)
void * openHandle; // the return value to clOpen
int bytesPerCore = 16; // how many bytes we want each core to process
int workItems = 32; // the total number workItems (threads sent to the Epiphany)
int i; // loop counter
cl_uchar * wrkArea1D; // the pointer to the malloc'd space (bytesPerCore * workItems)
cl_uchar * wrkArea2D;
// These variables will be useful when I get getDeviceInfo and getBuildInfo working
// unsigned int space; // the space required for clGetInfo style calls
// cl_build_status bOk; // the return value from clGetProgramBuildInfo
// size_t computeUnits; // return from GetDeviceInfo
// char strInfo[20];
FILE * pFile;
if(argc == 2)
clFile = argv[1];
else
clFile = kernFile;
pFile = fopen(clFile, "r");
if (pFile == NULL)
{
printf("Opening the Kernel file: %s produced an error(%d). Make sure that the source code variable kern has a valid path to the cl code and that the code is readable.\n", clFile, errno);
exit(0);
}
else
fclose(pFile); // only open the file to check that it is there and readable
debugReset(debugFile);
debugdebug(debugFile, (char*)"How many devices do we have?\n");
sprintf(msg, "About to malloc wrkArea1D: %d\n", workItems * bytesPerCore);
debugdebug(debugFile, msg);
wrkArea1D = (cl_uchar*) clmalloc(stdacc, workItems * bytesPerCore, 0);
wrkArea2D = (cl_uchar*) clmalloc(stdacc, workItems * bytesPerCore, 0);
for (i=0; i < workItems * bytesPerCore; i++)
wrkArea2D[i] = wrkArea1D[i] = 0;
sprintf(msg, "Well malloc worked! Opening kernel file:%s\n", clFile);
debugdebug(debugFile, msg);
openHandle = clopen(stdacc, clFile, CLLD_NOW);
// open the standard accellerator context (i.e. the Epiphany chip, reading in the .cl file and compiling it immediately
clndrange_t ndr1D = clndrange_init1d(NULL, // global offset (always zero)
((size_t)workItems), // total number of threads (get_global_id will return 0 to workItems
((size_t)bytesPerCore)); // How many bytes do we tell the kernel to process via get_local_size(0)
exKernel(openHandle, &ndr1D, (char*)"k_init1D", workItems, bytesPerCore, wrkArea1D, debugFile);
clndrange_t ndr2D = clndrange_init2d(NULL, // global offset (always zero)
((size_t)workItems), // total number of threads (get_global_id will return 0 to workItems
((size_t)(bytesPerCore/4)), // How many bytes do we tell the kernel to process via get_local_size(0)
NULL, // another useless global offset
((size_t)workItems), // a value that does not seem to do anything useful
((size_t)(bytesPerCore/4))); // How many rows to process per call returned by get_local_size(1)
exKernel(openHandle, &ndr2D, (char*)"k_init2D", workItems, bytesPerCore, wrkArea2D, debugFile);
// ============================================================================================================
// show the results
// ============================================================================================================
printf("The 1D data:\n");
for(i=0; i < workItems * bytesPerCore; i++)
{
printf("%u\t", wrkArea1D[i]);
if(((i+1) % bytesPerCore) == 0)
printf("\n");
}
printf("The 2D data:\n");
for(i=0; i < workItems * bytesPerCore; i++)
{
printf("%u\t", wrkArea2D[i]);
if(((i+1) % bytesPerCore) == 0)
printf("\n");
}
clfree(wrkArea1D);
clfree(wrkArea2D);
return 0;
}
int main()
{
cl_uint n = 64;
#if(1)
/* use default contexts, if no GPU use CPU */
CLCONTEXT* cp = (stdgpu)? stdgpu : stdcpu;
unsigned int devnum = 0;
void* clh = clopen(cp,"matvecmult.cl",CLLD_NOW);
cl_kernel krn = clsym(cp,clh,"matvecmult_kern",0);
/* allocate OpenCL device-sharable memory */
cl_float* aa = (float*)clmalloc(cp,n*n*sizeof(cl_float),0);
cl_float* b = (float*)clmalloc(cp,n*sizeof(cl_float),0);
cl_float* c = (float*)clmalloc(cp,n*sizeof(cl_float),0);
clndrange_t ndr = clndrange_init1d( 0, n, 64);
/* initialize vectors a[] and b[], zero c[] */
int i,j;
for(i=0;i<n;i++) for(j=0;j<n;j++) aa[i*n+j] = 1.1f*i*j;
for(i=0;i<n;i++) b[i] = 2.2f*i;
for(i=0;i<n;i++) c[i] = 0.0f;
/* define the computational domain and workgroup size */
//clndrange_t ndr = clndrange_init1d( 0, n, 64);
/* non-blocking sync vectors a and b to device memory (copy to GPU)*/
clmsync(cp,devnum,aa,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync(cp,devnum,b,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
/* set the kernel arguments */
clarg_set(cp,krn,0,n);
clarg_set_global(cp,krn,1,aa);
clarg_set_global(cp,krn,2,b);
clarg_set_global(cp,krn,3,c);
/* non-blocking fork of the OpenCL kernel to execute on the GPU */
clfork(cp,devnum,krn,&ndr,CL_EVENT_NOWAIT);
/* non-blocking sync vector c to host memory (copy back to host) */
clmsync(cp,0,c,CL_MEM_HOST|CL_EVENT_NOWAIT);
/* force execution of operations in command queue (non-blocking call) */
clflush(cp,devnum,0);
/* block on completion of operations in command queue */
clwait(cp,devnum,CL_ALL_EVENT);
for(i=0;i<n;i++) printf("%d %f %f\n",i,b[i],c[i]);
clfree(aa);
clfree(b);
clfree(c);
clclose(cp,clh);
#endif
}
int
ugenopen(dev_t dev, int flag, int mode, struct proc *p)
{
struct ugen_softc *sc;
int unit = UGENUNIT(dev);
int endpt = UGENENDPOINT(dev);
usb_endpoint_descriptor_t *edesc;
struct ugen_endpoint *sce;
int dir, isize;
usbd_status err;
struct usbd_xfer *xfer;
void *buf;
int i, j;
if (unit >= ugen_cd.cd_ndevs)
return (ENXIO);
sc = ugen_cd.cd_devs[unit];
if (sc == NULL)
return (ENXIO);
DPRINTFN(5, ("ugenopen: flag=%d, mode=%d, unit=%d endpt=%d\n",
flag, mode, unit, endpt));
if (sc == NULL || usbd_is_dying(sc->sc_udev))
return (ENXIO);
if (sc->sc_is_open[endpt])
return (EBUSY);
if (endpt == USB_CONTROL_ENDPOINT) {
sc->sc_is_open[USB_CONTROL_ENDPOINT] = 1;
return (0);
}
/* Make sure there are pipes for all directions. */
for (dir = OUT; dir <= IN; dir++) {
if (flag & (dir == OUT ? FWRITE : FREAD)) {
sce = &sc->sc_endpoints[endpt][dir];
if (sce == 0 || sce->edesc == 0)
return (ENXIO);
}
}
/* Actually open the pipes. */
/* XXX Should back out properly if it fails. */
for (dir = OUT; dir <= IN; dir++) {
if (!(flag & (dir == OUT ? FWRITE : FREAD)))
continue;
sce = &sc->sc_endpoints[endpt][dir];
sce->state = 0;
sce->timeout = USBD_NO_TIMEOUT;
DPRINTFN(5, ("ugenopen: sc=%p, endpt=%d, dir=%d, sce=%p\n",
sc, endpt, dir, sce));
edesc = sce->edesc;
switch (edesc->bmAttributes & UE_XFERTYPE) {
case UE_INTERRUPT:
if (dir == OUT) {
err = usbd_open_pipe(sce->iface,
edesc->bEndpointAddress, 0, &sce->pipeh);
if (err)
return (EIO);
break;
}
isize = UGETW(edesc->wMaxPacketSize);
if (isize == 0) /* shouldn't happen */
return (EINVAL);
sce->ibuf = malloc(isize, M_USBDEV, M_WAITOK);
DPRINTFN(5, ("ugenopen: intr endpt=%d,isize=%d\n",
endpt, isize));
clalloc(&sce->q, UGEN_IBSIZE, 0);
err = usbd_open_pipe_intr(sce->iface,
edesc->bEndpointAddress,
USBD_SHORT_XFER_OK, &sce->pipeh, sce,
sce->ibuf, isize, ugenintr,
USBD_DEFAULT_INTERVAL);
if (err) {
free(sce->ibuf, M_USBDEV, 0);
clfree(&sce->q);
return (EIO);
}
DPRINTFN(5, ("ugenopen: interrupt open done\n"));
break;
case UE_BULK:
err = usbd_open_pipe(sce->iface,
edesc->bEndpointAddress, 0, &sce->pipeh);
if (err)
return (EIO);
break;
case UE_ISOCHRONOUS:
if (dir == OUT)
return (EINVAL);
isize = UGETW(edesc->wMaxPacketSize);
if (isize == 0) /* shouldn't happen */
return (EINVAL);
sce->ibuf = malloc(isize * UGEN_NISOFRAMES,
M_USBDEV, M_WAITOK);
sce->cur = sce->fill = sce->ibuf;
sce->limit = sce->ibuf + isize * UGEN_NISOFRAMES;
DPRINTFN(5, ("ugenopen: isoc endpt=%d, isize=%d\n",
endpt, isize));
err = usbd_open_pipe(sce->iface,
edesc->bEndpointAddress, 0, &sce->pipeh);
if (err) {
free(sce->ibuf, M_USBDEV, 0);
return (EIO);
}
for(i = 0; i < UGEN_NISOREQS; ++i) {
sce->isoreqs[i].sce = sce;
xfer = usbd_alloc_xfer(sc->sc_udev);
if (xfer == 0)
goto bad;
sce->isoreqs[i].xfer = xfer;
buf = usbd_alloc_buffer
(xfer, isize * UGEN_NISORFRMS);
if (buf == 0) {
i++;
goto bad;
}
sce->isoreqs[i].dmabuf = buf;
for(j = 0; j < UGEN_NISORFRMS; ++j)
sce->isoreqs[i].sizes[j] = isize;
usbd_setup_isoc_xfer
(xfer, sce->pipeh, &sce->isoreqs[i],
sce->isoreqs[i].sizes,
UGEN_NISORFRMS, USBD_NO_COPY,
ugen_isoc_rintr);
(void)usbd_transfer(xfer);
}
DPRINTFN(5, ("ugenopen: isoc open done\n"));
break;
bad:
while (--i >= 0) /* implicit buffer free */
usbd_free_xfer(sce->isoreqs[i].xfer);
return (ENOMEM);
case UE_CONTROL:
sce->timeout = USBD_DEFAULT_TIMEOUT;
return (EINVAL);
}
}
sc->sc_is_open[endpt] = 1;
return (0);
}
int main()
{
cl_uint n = 1024;
/* use default contexts, if no GPU use CPU */
CLCONTEXT* cp = (stdgpu)? stdgpu : stdcpu;
unsigned int devnum = 0;
#ifdef __FreeBSD__
void* clh = clopen(cp,"matvecmult_special.cl",CLLD_NOW);
cl_kernel krn = clsym(cp,clh,"matvecmult_special_kern",0);
#else
cl_kernel krn = clsym(cp,0,"matvecmult_special_kern",0);
#endif
/* allocate OpenCL device-sharable memory */
cl_float* aa = (float*)clmalloc(cp,n*n*sizeof(cl_float),0);
cl_float* b = (float*)clmalloc(cp,n*sizeof(cl_float),0);
cl_float* c = (float*)clmalloc(cp,n*sizeof(cl_float),0);
/* initialize vectors a[] and b[], zero c[] */
int i,j;
for(i=0;i<n;i++) for(j=0;j<n;j++) aa[i*n+j] = 1.1f*i*j;
for(i=0;i<n;i++) b[i] = 2.2f*i;
for(i=0;i<n;i++) c[i] = 0.0f;
/***
*** Create a image2d allocation to be used as a read-only table.
*** The table will consist of a 24x24 array of float coefficients.
*** The clmctl() call is used to set the type and shape of the table.
*** Note that we will only use the first component of the float4 elements.
***/
cl_float4* table
= (cl_float4*)clmalloc(cp,24*24*sizeof(cl_float4),CL_MEM_DETACHED);
clmctl(table,CL_MCTL_SET_IMAGE2D,24,24,0);
clmattach(cp,table);
/* initialize the table to some contrived values */
for(i=0;i<24;i++) for(j=0;j<24;j++) table[i*24+j].x = 0.125f*(i-j);
/* define the computational domain and workgroup size */
clndrange_t ndr = clndrange_init1d( 0, n, 64);
/* non-blocking sync vectors a and b to device memory (copy to GPU)*/
clmsync(cp,devnum,aa,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync(cp,devnum,b,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync(cp,devnum,table,CL_MEM_DEVICE|CL_EVENT_NOWAIT);
/* set the kernel arguments */
clarg_set(cp,krn,0,n);
clarg_set_global(cp,krn,1,aa);
clarg_set_global(cp,krn,2,b);
clarg_set_global(cp,krn,3,c);
clarg_set_global(cp,krn,4,table);
/* non-blocking fork of the OpenCL kernel to execute on the GPU */
clfork(cp,devnum,krn,&ndr,CL_EVENT_NOWAIT);
/* non-blocking sync vector c to host memory (copy back to host) */
clmsync(cp,0,c,CL_MEM_HOST|CL_EVENT_NOWAIT);
/* block on completion of operations in command queue */
clwait(cp,devnum,CL_ALL_EVENT);
for(i=0;i<n;i++) printf("%d %f %f\n",i,b[i],c[i]);
clfree(aa);
clfree(b);
clfree(c);
#ifdef __FreeBSD__
clclose(cp,clh);
#endif
}
int main()
{
cl_uint n = 1024;
/* use default contexts, if no ACCELERATOR use CPU */
CLCONTEXT* cp = (stdacc)? stdacc : stdcpu;
unsigned int devnum = 0;
/******************************************************************
*** this example requires the .cl file to be available at run-time
*** and shows how to pass compiler options to the OCL compiler
******************************************************************/
void* clh = clopen(cp,"outerprod.cl",CLLD_NOBUILD);
clbuild(cp,clh,"-D COEF=2", 0);
cl_kernel krn = clsym(cp,clh,"outerprod_kern",0);
if (!krn) { fprintf(stderr,"error: no OpenCL kernel\n"); exit(-1); }
/* allocate OpenCL device-sharable memory */
cl_float* a = (float*)clmalloc(cp,n*sizeof(cl_float),0);
cl_float* b = (float*)clmalloc(cp,n*sizeof(cl_float),0);
cl_float* c = (float*)clmalloc(cp,n*sizeof(cl_float),0);
/* initialize vectors a[] and b[], zero c[] */
int i;
for(i=0;i<n;i++) a[i] = 1.1f*i;
for(i=0;i<n;i++) b[i] = 2.2f*i;
for(i=0;i<n;i++) c[i] = 0.0f;
/* non-blocking sync vectors a and b to device memory (copy to GPU)*/
clmsync(cp,devnum,a,CL_MEM_DEVICE|CL_EVENT_WAIT);
clmsync(cp,devnum,b,CL_MEM_DEVICE|CL_EVENT_WAIT);
/* define the computational domain and workgroup size */
clndrange_t ndr = clndrange_init1d( 0, n, 16);
/* set the kernel arguments */
clarg_set_global(cp,krn,0,a);
clarg_set_global(cp,krn,1,b);
clarg_set_global(cp,krn,2,c);
/* non-blocking fork of the OpenCL kernel to execute on the GPU */
clfork(cp,devnum,krn,&ndr,CL_EVENT_NOWAIT);
/* block on completion of operations in command queue */
clwait(cp,devnum,CL_ALL_EVENT);
/* non-blocking sync vector c to host memory (copy back to host) */
clmsync(cp,0,c,CL_MEM_HOST|CL_EVENT_NOWAIT);
/* block on completion of operations in command queue */
clwait(cp,devnum,CL_ALL_EVENT);
for(i=0;i<n;i++) printf("%d %f %f %f\n",i,a[i],b[i],c[i]);
clfree(a);
clfree(b);
clfree(c);
clclose(cp,clh);
}
/*
* Calculates the path loss matrix for many transmitters and saves the
* output to one common raster map.
*
* int nrows .............. the number of rows in the DEM.
* int ncols .............. the number of columns in the DEM.
* int tx_coord_col ....... transmitter column coordinate (in raster cells).
* int tx_coord_row ....... transmitter row coordinate (in raster cells).
* int tx_elevation ....... transmitter height above sea level.
* int anthena_height ..... height of transmitter anthena above ground.
* float rx_height ........ receiver height above ground.
* float frequency ........ transmitter frequency in Mhz.
* int radius ............. calculation radius around to the transmitter
* (in raster cells).
* int infd ............... file descriptor to read from the DEM.
* int outfd .............. file descriptor to write resulting raster.
* RASTER_MAP_TYPE rtype .. DEM element data type.
*
*/
void hata_cl (const int nrows, const int ncols,
const int tx_coord_col, const int tx_coord_row,
const int tx_elevation, const int tx_anthena_height,
const float rx_height, const float frequency,
const int radius,
const int infd, const int outfd,
RASTER_MAP_TYPE rtype)
{
int i;
/* number of processing tiles needed around each transmitter */
int ntile = (radius*2) / _HATA_WITEM_PER_DIM_ + 1;
G_message ("Receiver height above ground is %5.2f mts", rx_height);
G_message ("Transmitter frequency is %7.2f Mhz", frequency);
G_message ("Calculation radius is %d raster cells", radius);
CONTEXT * ctx = init_context ( );
/* initialize raster data */
void *dem_in = load_raster (ctx, nrows, ncols, infd, rtype);
void *dem_out = null_raster (ctx, nrows, ncols, rtype, NULL);
/* load, compile and link the OpenCL kernel file */
void* h = clopen (ctx, NULL, CLLD_NOW);
/* select a kernel function from the loaded file */
cl_kernel krn = clsym (ctx, h, "hata_urban_kern", CLLD_NOW);
/* defines a 2D matrix of work groups & work items */
size_t global_wsize = ntile * _HATA_WITEM_PER_DIM_;
size_t local_wsize = _HATA_WITEM_PER_DIM_;
if ((global_wsize % local_wsize) != 0)
G_fatal_error ("Work group dimensions are incorrect.");
else
G_message ("Creating %d work-group(s) of [%d x %d] threads each",
global_wsize/local_wsize,
local_wsize, local_wsize);
clndrange_t ndr = clndrange_init2d (0,global_wsize,local_wsize,
0,global_wsize,local_wsize);
// correction factor ahr
float ahr = (1.1f*log10(frequency) - 0.7f)*rx_height -
(1.56f*log10(frequency) - 0.8f);
/* set kernel parameters and local memory size */
clarg_set (ctx, krn, 0, ncols);
clarg_set (ctx, krn, 3, rx_height);
clarg_set (ctx, krn, 4, frequency);
clarg_set (ctx, krn, 5, ahr);
size_t lmem_size = _HATA_WITEM_PER_DIM_*_HATA_WITEM_PER_DIM_*
sizeof(cl_float2);
if (lmem_size > _HATA_MAX_LOCAL_MEM_)
G_fatal_error ("Allocated local memory (%d) exceeds %d bytes.", lmem_size,
_HATA_MAX_LOCAL_MEM_);
clarg_set_local (ctx, krn, 8, lmem_size);
/* transfer data to the device */
clmsync (ctx, 0, dem_in, CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync (ctx, 0, dem_out, CL_MEM_DEVICE|CL_EVENT_NOWAIT);
/* set remaining kernel parameters */
clarg_set_global (ctx, krn, 6, dem_in);
clarg_set_global (ctx, krn, 7, dem_out);
G_message ("Simulating ...");
/* test transmitters
ASNEBE | 1468 | 1200
ASKZK | 1395 | 1180
ASPAR | 1400 | 1245
ASMARG | 1460 | 1203
ASTARA | 1422 | 1235
ASKZ | 1396 | 1181
ASENTP | 1387 | 1185
ASISKP | 1399 | 1189
ASEME | 1448 | 1262
ASOSTR | 1498 | 1253
ASMELT | 1429 | 1196
ASMART | 1446 | 1213
ASLOM | 1425 | 1226
ASAZU | 1421 | 1235
ASONCE | 1433 | 1208
ASMAR | 1404 | 1150
ASTEP | 1450 | 1235
ASAVLJ | 1417 | 1182
ASTOZI | 1436 | 1198
ASTRAL | 1422 | 1227
ASTEG | 1404 | 1193
*/
const int ntest_tx = 21;
cl_int2 *test_tx = malloc (ntest_tx * sizeof(cl_int2));
test_tx[0] = (cl_int2) {1468,1200};
test_tx[1] = (cl_int2) {1395,1180};
test_tx[2] = (cl_int2) {1400,1245};
test_tx[3] = (cl_int2) {1460,1203};
test_tx[4] = (cl_int2) {1422,1235};
test_tx[5] = (cl_int2) {1396,1181};
test_tx[6] = (cl_int2) {1387,1185};
test_tx[7] = (cl_int2) {1399,1189};
test_tx[8] = (cl_int2) {1448,1262};
test_tx[9] = (cl_int2) {1498,1253};
test_tx[10] = (cl_int2) {1429,1196};
test_tx[11] = (cl_int2) {1446,1213};
test_tx[12] = (cl_int2) {1425,1226};
test_tx[13] = (cl_int2) {1421,1235};
test_tx[14] = (cl_int2) {1433,1208};
test_tx[15] = (cl_int2) {1404,1150};
test_tx[16] = (cl_int2) {1450,1235};
test_tx[17] = (cl_int2) {1417,1182};
test_tx[18] = (cl_int2) {1436,1198};
test_tx[19] = (cl_int2) {1422,1227};
test_tx[20] = (cl_int2) {1404,1193};
/* for each transmitter */
for (i = 0; i < ntest_tx; i++)
{
/* display completion percentage */
G_percent (i, ntest_tx, 1);
/* Transmitter coordinates */
cl_int2 tx_coord = test_tx[i];
//G_message ("The transmitter is located at %d, %d", tx_coord.x,
// tx_coord.y);
/* Transmitter heights */
//G_message ("Transmitter height above sea level is %d mts", tx_elevation);
//G_message ("Transmitter anthena height above ground is %d mts", tx_anthena_height);
/* transmitter data */
cl_int4 tx_data = (cl_int4) {tx_coord.x, tx_coord.y,
tx_elevation, tx_anthena_height};
/* tile offset within the raster */
cl_int2 tx_offset = (cl_int2) {tx_coord.x - radius,
tx_coord.y - radius};
/* set (more) remaining kernel parameters */
clarg_set (ctx, krn, 1, tx_data);
clarg_set (ctx, krn, 2, tx_offset);
/* start kernel execution */
clfork (ctx, 0, krn, &ndr, CL_EVENT_NOWAIT);
clwait (ctx, 0, CL_KERNEL_EVENT|CL_MEM_EVENT|CL_EVENT_RELEASE);
}
/* sync memory */
clmsync (ctx, 0, dem_out, CL_MEM_HOST|CL_EVENT_WAIT);
/* write the calculation output */
save_raster (dem_out, nrows, ncols, outfd);
/* free allocated resources */
free (test_tx);
clclose (ctx, h);
clfree (dem_out);
clfree (dem_in);
}
/*
* Calculates an interference matrix resulting from the tx power of all
* transmitters combined with their respective path losses and the thermal
* noise. The output is saved in one raster map.
*
* int nrows .............. the number of rows in the DEM.
* int ncols .............. the number of columns in the DEM.
* int tx_coord_col ....... transmitter column coordinate (in raster cells).
* int tx_coord_row ....... transmitter row coordinate (in raster cells).
* int tx_elevation ....... transmitter height above sea level.
* int anthena_height ..... height of transmitter anthena above ground.
* float rx_height ........ receiver height above ground.
* float frequency ........ transmitter frequency in Mhz.
* int radius ............. calculation radius around to the transmitter
* (in raster cells).
* int infd ............... file descriptor to read from the DEM.
* int outfd .............. file descriptor to write the interference
* data to.
* RASTER_MAP_TYPE rtype .. DEM element data type.
*
*/
void hata_interference_cl (const int nrows, const int ncols,
const int tx_coord_col, const int tx_coord_row,
const int tx_elevation, const int tx_anthena_height,
const float rx_height, const float frequency,
const int radius,
const int infd, const int outfd,
RASTER_MAP_TYPE rtype)
{
int i;
/* number of processing tiles needed around each transmitter */
int ntile = (radius*2) / _HATA_WITEM_PER_DIM_ + 1;
G_message ("Receiver height above ground is %5.2f mts", rx_height);
G_message ("Transmitter frequency is %7.2f Mhz", frequency);
G_message ("Calculation radius is %d raster cells", radius);
CONTEXT * ctx = init_context ( );
/* initialize raster data */
void *dem_in = load_raster (ctx, nrows, ncols, infd, rtype);
/* interference matrix */
void *qrm_out = alloc_raster (ctx, nrows, ncols, rtype,
(FCELL)_HATA_THERMAL_NOISE_, NULL);
/* load, compile and link the OpenCL kernel file */
void* h = clopen (ctx, NULL, CLLD_NOW);
/* select a kernel function from the loaded file */
cl_kernel krn = clsym (ctx, h, "hata_urban_interference", CLLD_NOW);
/* defines a 2D matrix of work groups & work items */
size_t global_wsize = ntile * _HATA_WITEM_PER_DIM_;
size_t local_wsize = _HATA_WITEM_PER_DIM_;
if ((global_wsize % local_wsize) != 0)
G_fatal_error ("Work group dimensions are incorrect.");
else
G_message ("Creating %d work-group(s) of [%d x %d] threads each",
global_wsize/local_wsize,
local_wsize, local_wsize);
clndrange_t ndr = clndrange_init2d (0,global_wsize,local_wsize,
0,global_wsize,local_wsize);
// correction factor ahr
float ahr = (1.1f*log10(frequency) - 0.7f)*rx_height -
(1.56f*log10(frequency) - 0.8f);
/* set kernel parameters and local memory size */
clarg_set (ctx, krn, 0, ncols);
clarg_set (ctx, krn, 3, rx_height);
clarg_set (ctx, krn, 4, frequency);
clarg_set (ctx, krn, 5, ahr);
size_t lmem_size = _HATA_WITEM_PER_DIM_*_HATA_WITEM_PER_DIM_*
sizeof(cl_float2);
if (lmem_size > _HATA_MAX_LOCAL_MEM_)
G_fatal_error ("Allocated local memory (%d) exceeds %d bytes.", lmem_size,
_HATA_MAX_LOCAL_MEM_);
clarg_set_local (ctx, krn, 8, lmem_size);
/* transfer data to the device */
clmsync (ctx, 0, dem_in, CL_MEM_DEVICE|CL_EVENT_NOWAIT);
clmsync (ctx, 0, qrm_out, CL_MEM_DEVICE|CL_EVENT_WAIT);
/* set remaining kernel parameters */
clarg_set_global (ctx, krn, 6, dem_in);
clarg_set_global (ctx, krn, 7, qrm_out);
G_message ("Simulating ...");
/**
* Test transmitters
*
* ASNEBE | 1468 | 1200
* ASKZK | 1395 | 1180
* ALEK | 1408 | 1211
* ANDREJ | 1431 | 1212
* AMALENA | 1462 | 1272
* APOLJE | 1483 | 1227
* AMONS | 1381 | 1230
* ANTENA | 1397 | 1238
* ATOMAC | 1443 | 1194
* ATUNEL | 1424 | 1239
* AKOSET | 1398 | 1206
* ASPAR | 1400 | 1245
* ACHEMO | 1431 | 1226
* ALIVAD | 1423 | 1248
* ASMARG | 1460 | 1203
* ABTC | 1451 | 1215
* AZIMA | 1471 | 1236
* ASTARA | 1422 | 1235
* AVEGAS | 1434 | 1257
* APOVSE | 1436 | 1232
* ABTCST | 1449 | 1213
* AMOBI | 1429 | 1222
* APODUT | 1378 | 1202
* AOBZEL | 1444 | 1222
* AGRIC | 1385 | 1207
* ADELO | 1421 | 1222
* AJEZA | 1459 | 1182
* ASKZ | 1396 | 1181
* ATLK | 1422 | 1225
* AGR | 1424 | 1220
* ADNEVN | 1426 | 1232
* AMHOTE | 1409 | 1212
* AVIC | 1414 | 1236
* AJOZEF | 1429 | 1234
* ATVSLO | 1425 | 1228
* ACRNUC | 1445 | 1172
* ANGEL | 1508 | 1219
* AKOSEG | 1401 | 1210
* AMSC | 1429 | 1222
* AGRAM | 1421 | 1195
* ARAKTK | 1436 | 1249
* AIMKO | 1445 | 1179
* AKOVIN | 1412 | 1214
* ADOLGI | 1388 | 1250
* APETER | 1431 | 1231
* ASENTP | 1387 | 1185
* AOBI | 1443 | 1268
* AMZS | 1427 | 1203
* AVICTK | 1403 | 1240
* AZUR | 1397 | 1238
* AKAMNA | 1388 | 1200
* APOLCE | 1482 | 1228
* AKASEL | 1497 | 1232
* ASISKP | 1399 | 1189
* AKOZAR | 1373 | 1247
* ASEME | 1448 | 1262
* AURSKA | 1424 | 1220
* ABEZIT | 1427 | 1199
* AVEROV | 1416 | 1201
* ATEGR | 1419 | 1209
* ATRUB | 1423 | 1231
* AKOLIN | 1436 | 1221
* AVELAN | 1439 | 1218
* ABEZI | 1431 | 1216
* AGAMGD | 1418 | 1148
* AGURS | 1431 | 1237
* AGPG | 1419 | 1237
* AKOSME | 1397 | 1210
* AZALOG | 1500 | 1225
* ASOSTR | 1498 | 1253
* AROZNA | 1404 | 1235
* AVRHO | 1380 | 1242
* ARAK | 1430 | 1244
* AVILA | 1404 | 1230
* AKOLEZ | 1413 | 1241
* AJEZKZ | 1427 | 1185
* AVIZMA | 1386 | 1173
* AKOTNI | 1429 | 1226
* AMOSTE | 1443 | 1227
* ATOPNI | 1425 | 1212
* AGZ | 1429 | 1209
* APRZAC | 1386 | 1192
* ALITIJ | 1455 | 1235
* ATEHNO | 1387 | 1233
* ASMELT | 1429 | 1196
* ACRGMA | 1439 | 1166
* AFORD | 1439 | 1215
* ADRAVC | 1396 | 1197
* ABROD | 1388 | 1168
* ATABOR | 1429 | 1228
* ATRNOV | 1421 | 1242
* AGOLOV | 1467 | 1263
* ARESEV | 1438 | 1228
* ABTCMI | 1450 | 1216
* ASMART | 1446 | 1213
* APROGA | 1382 | 1251
* ABRDO | 1382 | 1235
* AKONEX | 1387 | 1237
* AZITO | 1456 | 1213
* ASLOM | 1425 | 1226
* APLAMA | 1448 | 1202
* AGRAD | 1470 | 1218
* AFUZIN | 1463 | 1228
* ASAZU | 1421 | 1235
* ATOTRA | 1463 | 1233
* AHALA | 1414 | 1221
* ASONCE | 1433 | 1208
* ALMLEK | 1422 | 1202
* ACERNE | 1415 | 1212
* ADRAV | 1396 | 1201
* ALGRAD | 1424 | 1234
* ALPP | 1406 | 1205
* ANAMA | 1420 | 1231
* ATIVO | 1418 | 1226
* AZAPUZ | 1392 | 1196
* ASMAR | 1404 | 1150
* ACANK | 1416 | 1233
* ATACGD | 1390 | 1156
* APOPTV | 1435 | 1211
* AFRANK | 1417 | 1220
* ASTEP | 1450 | 1235
* AEJATA | 1445 | 1218
* AZADOB | 1482 | 1206
* AELNA | 1436 | 1177
* AZELEZ | 1428 | 1218
* AMURGL | 1415 | 1246
* ACIGO | 1423 | 1225
* AIJS | 1407 | 1241
* AELOK | 1459 | 1225
* ATACBR | 1385 | 1160
* AJELSA | 1416 | 1259
* AVRHOV | 1394 | 1242
* ABIZOV | 1472 | 1247
* ACITYP | 1454 | 1214
* ASAVLJ | 1417 | 1182
* AKLINI | 1434 | 1228
* AKODEL | 1439 | 1236
* ASTOZI | 1436 | 1198
* AKORA | 1422 | 1226
* AHLEV | 1388 | 1163
* ATOPLA | 1451 | 1224
* AMALEN | 1461 | 1273
* AVECNA | 1391 | 1222
* ANADGO | 1462 | 1178
* ABOKAL | 1379 | 1231
* AVARNO | 1404 | 1250
* ASTRAL | 1422 | 1227
* AVEVCE | 1488 | 1234
* ARNC | 1423 | 1225
* AGEOPL | 1406 | 1198
* AMOBLI | 1415 | 1200
* AVLADA | 1415 | 1234
* ASTEG | 1404 | 1193
* AKOSEZ | 1392 | 1207
*/
const int ntest_tx = 154;
cl_int2 *test_tx = malloc (ntest_tx * sizeof(cl_int2));
test_tx[0] = (cl_int2) {1468,1200};
test_tx[1] = (cl_int2) {1395,1180};
test_tx[2] = (cl_int2) {1408,1211};
test_tx[3] = (cl_int2) {1431,1212};
test_tx[4] = (cl_int2) {1462,1272};
test_tx[5] = (cl_int2) {1483,1227};
test_tx[6] = (cl_int2) {1381,1230};
test_tx[7] = (cl_int2) {1397,1238};
test_tx[8] = (cl_int2) {1443,1194};
test_tx[9] = (cl_int2) {1424,1239};
test_tx[10] = (cl_int2) {1398,1206};
test_tx[11] = (cl_int2) {1400,1245};
test_tx[12] = (cl_int2) {1431,1226};
test_tx[13] = (cl_int2) {1423,1248};
test_tx[14] = (cl_int2) {1460,1203};
test_tx[15] = (cl_int2) {1451,1215};
test_tx[16] = (cl_int2) {1471,1236};
test_tx[17] = (cl_int2) {1422,1235};
test_tx[18] = (cl_int2) {1434,1257};
test_tx[19] = (cl_int2) {1436,1232};
test_tx[20] = (cl_int2) {1449,1213};
test_tx[21] = (cl_int2) {1429,1222};
test_tx[22] = (cl_int2) {1378,1202};
test_tx[23] = (cl_int2) {1444,1222};
test_tx[24] = (cl_int2) {1385,1207};
test_tx[25] = (cl_int2) {1421,1222};
test_tx[26] = (cl_int2) {1459,1182};
test_tx[27] = (cl_int2) {1396,1181};
test_tx[28] = (cl_int2) {1422,1225};
test_tx[29] = (cl_int2) {1424,1220};
test_tx[30] = (cl_int2) {1426,1232};
test_tx[31] = (cl_int2) {1409,1212};
test_tx[32] = (cl_int2) {1414,1236};
test_tx[33] = (cl_int2) {1429,1234};
test_tx[34] = (cl_int2) {1425,1228};
test_tx[35] = (cl_int2) {1445,1172};
test_tx[36] = (cl_int2) {1508,1219};
test_tx[37] = (cl_int2) {1401,1210};
test_tx[38] = (cl_int2) {1429,1222};
test_tx[39] = (cl_int2) {1421,1195};
test_tx[40] = (cl_int2) {1436,1249};
test_tx[41] = (cl_int2) {1445,1179};
test_tx[42] = (cl_int2) {1412,1214};
test_tx[43] = (cl_int2) {1388,1250};
test_tx[44] = (cl_int2) {1431,1231};
test_tx[45] = (cl_int2) {1387,1185};
test_tx[46] = (cl_int2) {1443,1268};
test_tx[47] = (cl_int2) {1427,1203};
test_tx[48] = (cl_int2) {1403,1240};
test_tx[49] = (cl_int2) {1397,1238};
test_tx[50] = (cl_int2) {1388,1200};
test_tx[51] = (cl_int2) {1482,1228};
test_tx[52] = (cl_int2) {1497,1232};
test_tx[53] = (cl_int2) {1399,1189};
test_tx[54] = (cl_int2) {1373,1247};
test_tx[55] = (cl_int2) {1448,1262};
test_tx[56] = (cl_int2) {1424,1220};
test_tx[57] = (cl_int2) {1427,1199};
test_tx[58] = (cl_int2) {1416,1201};
test_tx[59] = (cl_int2) {1419,1209};
test_tx[60] = (cl_int2) {1423,1231};
test_tx[61] = (cl_int2) {1436,1221};
test_tx[62] = (cl_int2) {1439,1218};
test_tx[63] = (cl_int2) {1431,1216};
test_tx[64] = (cl_int2) {1418,1148};
test_tx[65] = (cl_int2) {1431,1237};
test_tx[66] = (cl_int2) {1419,1237};
test_tx[67] = (cl_int2) {1397,1210};
test_tx[68] = (cl_int2) {1500,1225};
test_tx[69] = (cl_int2) {1498,1253};
test_tx[70] = (cl_int2) {1404,1235};
test_tx[71] = (cl_int2) {1380,1242};
test_tx[72] = (cl_int2) {1430,1244};
test_tx[73] = (cl_int2) {1404,1230};
test_tx[74] = (cl_int2) {1413,1241};
test_tx[75] = (cl_int2) {1427,1185};
test_tx[76] = (cl_int2) {1386,1173};
test_tx[77] = (cl_int2) {1429,1226};
test_tx[78] = (cl_int2) {1443,1227};
test_tx[79] = (cl_int2) {1425,1212};
test_tx[80] = (cl_int2) {1429,1209};
test_tx[81] = (cl_int2) {1386,1192};
test_tx[82] = (cl_int2) {1455,1235};
test_tx[83] = (cl_int2) {1387,1233};
test_tx[84] = (cl_int2) {1429,1196};
test_tx[85] = (cl_int2) {1439,1166};
test_tx[86] = (cl_int2) {1439,1215};
test_tx[87] = (cl_int2) {1396,1197};
test_tx[88] = (cl_int2) {1388,1168};
test_tx[89] = (cl_int2) {1429,1228};
test_tx[90] = (cl_int2) {1421,1242};
test_tx[91] = (cl_int2) {1467,1263};
test_tx[92] = (cl_int2) {1438,1228};
test_tx[93] = (cl_int2) {1450,1216};
test_tx[94] = (cl_int2) {1446,1213};
test_tx[95] = (cl_int2) {1382,1251};
test_tx[96] = (cl_int2) {1382,1235};
test_tx[97] = (cl_int2) {1387,1237};
test_tx[98] = (cl_int2) {1456,1213};
test_tx[99] = (cl_int2) {1425,1226};
test_tx[100] = (cl_int2) {1448,1202};
test_tx[101] = (cl_int2) {1470,1218};
test_tx[102] = (cl_int2) {1463,1228};
test_tx[103] = (cl_int2) {1421,1235};
test_tx[104] = (cl_int2) {1463,1233};
test_tx[105] = (cl_int2) {1414,1221};
test_tx[106] = (cl_int2) {1433,1208};
test_tx[107] = (cl_int2) {1422,1202};
test_tx[108] = (cl_int2) {1415,1212};
test_tx[109] = (cl_int2) {1396,1201};
test_tx[110] = (cl_int2) {1424,1234};
test_tx[111] = (cl_int2) {1406,1205};
test_tx[112] = (cl_int2) {1420,1231};
test_tx[113] = (cl_int2) {1418,1226};
test_tx[114] = (cl_int2) {1392,1196};
test_tx[115] = (cl_int2) {1404,1150};
test_tx[116] = (cl_int2) {1416,1233};
test_tx[117] = (cl_int2) {1390,1156};
test_tx[118] = (cl_int2) {1435,1211};
test_tx[119] = (cl_int2) {1417,1220};
test_tx[120] = (cl_int2) {1450,1235};
test_tx[121] = (cl_int2) {1445,1218};
test_tx[122] = (cl_int2) {1482,1206};
test_tx[123] = (cl_int2) {1436,1177};
test_tx[124] = (cl_int2) {1428,1218};
test_tx[125] = (cl_int2) {1415,1246};
test_tx[126] = (cl_int2) {1423,1225};
test_tx[127] = (cl_int2) {1407,1241};
test_tx[128] = (cl_int2) {1459,1225};
test_tx[129] = (cl_int2) {1385,1160};
test_tx[130] = (cl_int2) {1416,1259};
test_tx[131] = (cl_int2) {1394,1242};
test_tx[132] = (cl_int2) {1472,1247};
test_tx[133] = (cl_int2) {1454,1214};
test_tx[134] = (cl_int2) {1417,1182};
test_tx[135] = (cl_int2) {1434,1228};
test_tx[136] = (cl_int2) {1439,1236};
test_tx[137] = (cl_int2) {1436,1198};
test_tx[138] = (cl_int2) {1422,1226};
test_tx[139] = (cl_int2) {1388,1163};
test_tx[140] = (cl_int2) {1451,1224};
test_tx[141] = (cl_int2) {1461,1273};
test_tx[142] = (cl_int2) {1391,1222};
test_tx[143] = (cl_int2) {1462,1178};
test_tx[144] = (cl_int2) {1379,1231};
test_tx[145] = (cl_int2) {1404,1250};
test_tx[146] = (cl_int2) {1422,1227};
test_tx[147] = (cl_int2) {1488,1234};
test_tx[148] = (cl_int2) {1423,1225};
test_tx[149] = (cl_int2) {1406,1198};
test_tx[150] = (cl_int2) {1415,1200};
test_tx[151] = (cl_int2) {1415,1234};
test_tx[152] = (cl_int2) {1404,1193};
test_tx[153] = (cl_int2) {1392,1207};
/* for each transmitter */
for (i = 0; i < ntest_tx; i++)
{
/* display completion percentage */
G_percent (i, ntest_tx, 1);
/* Transmitter coordinates */
cl_int2 tx_coord = test_tx[i];
G_message ("The transmitter is located at %d, %d", tx_coord.x,
tx_coord.y);
/* Transmitter heights */
//G_message ("Transmitter height above sea level is %d mts", tx_elevation);
//G_message ("Transmitter anthena height above ground is %d mts", tx_anthena_height);
/* transmitter data */
cl_int4 tx_data = (cl_int4) {tx_coord.x, tx_coord.y,
tx_elevation, tx_anthena_height};
/* tile offset within the raster */
cl_int2 tx_offset = (cl_int2) {tx_coord.x - radius,
tx_coord.y - radius};
/* set (more) remaining kernel parameters */
clarg_set (ctx, krn, 1, tx_data);
clarg_set (ctx, krn, 2, tx_offset);
/* start kernel execution */
clfork (ctx, 0, krn, &ndr, CL_EVENT_NOWAIT);
clwait (ctx, 0, CL_KERNEL_EVENT|CL_MEM_EVENT|CL_EVENT_RELEASE);
}
/* sync memory */
clmsync (ctx, 0, qrm_out, CL_MEM_HOST|CL_EVENT_WAIT);
/* write the calculation output */
save_raster (qrm_out, nrows, ncols, outfd);
/* free allocated resources */
free (test_tx);
clclose (ctx, h);
clfree (qrm_out);
clfree (dem_in);
}