void Ableton::update()
{
while(receiver.hasWaitingMessages())
{
ofxOscMessage m;
receiver.getNextMessage(&m);
displayOscMessage(m);
if (m.getAddress() == "/live/device/param") {
getParameterUpdate(m);
}
else if (m.getAddress() == "/live/devicelist") {
getDeviceList(m);
}
else if (m.getAddress() == "/live/device/allparam") {
getDeviceParameters(m);
}
else if (m.getAddress() == "/live/device/range") {
getDeviceParametersRange(m);
}
else if (m.getAddress() == "/live/scenes") {
getNumScenes(m);
}
else if (m.getAddress() == "/live/tracks") {
getNumTracks(m);
}
else {
displayOscMessage(m);
}
}
}
int VideoCaptureV4L2::listDevices() {
std::vector<LinuxCaptureDevice> devices = getDeviceList();
if(!devices.size()) {
printf("WARNING: no video4linux devices found\n");
return 0;
}
struct v4l2_input inputs[V4L2_MAX_QUERYABLE_INPUTS];
int c = 0;
for(std::vector<LinuxCaptureDevice>::iterator it = devices.begin(); it != devices.end(); ++it) {
printf("[%d] %s, (idVendor: %s, idProduct: %s)\n", c++, (*it).path.c_str(), (*it).id_vendor.c_str(), (*it).id_product.c_str());
}
return devices.size();
}
//------------------------------------------------------------
vector<ofSoundDevice> ofSoundStream::getMatchingDevices(const std::string& name, unsigned int inChannels, unsigned int outChannels) const {
vector<ofSoundDevice> devs = getDeviceList();
vector<ofSoundDevice> hits;
for(size_t i = 0; i < devs.size(); i++) {
bool nameMatch = devs[i].name.find(name) != string::npos;
bool inMatch = (inChannels == UINT_MAX) || (devs[i].inputChannels == inChannels);
bool outMatch = (outChannels == UINT_MAX) || (devs[i].outputChannels == outChannels);
if(nameMatch && inMatch && outMatch) {
hits.push_back(devs[i]);
}
}
return hits;
}
QString CameraDevice::getDefaultDeviceName()
{
QString defaultdev = Settings::getInstance().getVideoDev();
if (!getDefaultInputFormat())
return defaultdev;
QVector<QPair<QString, QString>> devlist = getDeviceList();
for (const QPair<QString,QString>& device : devlist)
if (defaultdev == device.first)
return defaultdev;
if (devlist.isEmpty())
return defaultdev;
return devlist[0].first;
}
//--------------------------------------------------------------------------------
bool ofxKinectV2::open(unsigned int deviceId){
vector <KinectDeviceInfo> devices = getDeviceList();
if( devices.size() == 0 ){
ofLogError("ofxKinectV2::open") << "no devices connected!";
return false;
}
if( deviceId >= devices.size() ){
ofLogError("ofxKinectV2::open") << " deviceId " << deviceId << " is bigger or equal to the number of connected devices " << devices.size() << endl;
return false;
}
string serial = devices[deviceId].serial;
return open(serial);
}
int main(int argc, char *argv[])
{
float *h_psum; // vector to hold partial sum
int in_nsteps = INSTEPS; // default number of steps (updated later to device prefereable)
int niters = ITERS; // number of iterations
int nsteps;
float step_size;
::size_t nwork_groups;
::size_t max_size, work_group_size = 8;
float pi_res;
cl::Buffer d_partial_sums;
try
{
cl_uint deviceIndex = 0;
parseArguments(argc, argv, &deviceIndex);
// Get list of devices
std::vector<cl::Device> devices;
unsigned numDevices = getDeviceList(devices);
// Check device index in range
if (deviceIndex >= numDevices)
{
std::cout << "Invalid device index (try '--list')\n";
return EXIT_FAILURE;
}
cl::Device device = devices[deviceIndex];
std::string name;
getDeviceName(device, name);
std::cout << "\nUsing OpenCL device: " << name << "\n";
std::vector<cl::Device> chosen_device;
chosen_device.push_back(device);
cl::Context context(chosen_device);
cl::CommandQueue queue(context, device);
// Create the program object
cl::Program program(context, util::loadProgram("../pi_ocl.cl"), true);
// Create the kernel object for quering information
cl::Kernel ko_pi(program, "pi");
// Get the work group size
work_group_size = ko_pi.getWorkGroupInfo<CL_KERNEL_WORK_GROUP_SIZE>(device);
//printf("wgroup_size = %lu\n", work_group_size);
cl::make_kernel<int, float, cl::LocalSpaceArg, cl::Buffer> pi(program, "pi");
// Now that we know the size of the work_groups, we can set the number of work
// groups, the actual number of steps, and the step size
nwork_groups = in_nsteps/(work_group_size*niters);
if ( nwork_groups < 1) {
nwork_groups = device.getInfo<CL_DEVICE_MAX_COMPUTE_UNITS>();
work_group_size=in_nsteps / (nwork_groups*niters);
}
nsteps = work_group_size * niters * nwork_groups;
step_size = 1.0f/static_cast<float>(nsteps);
std::vector<float> h_psum(nwork_groups);
printf(
" %d work groups of size %d. %d Integration steps\n",
(int)nwork_groups,
(int)work_group_size,
nsteps);
d_partial_sums = cl::Buffer(context, CL_MEM_WRITE_ONLY, sizeof(float) * nwork_groups);
util::Timer timer;
// Execute the kernel over the entire range of our 1d input data set
// using the maximum number of work group items for this device
pi(
cl::EnqueueArgs(
queue,
cl::NDRange(nsteps / niters),
cl::NDRange(work_group_size)),
niters,
step_size,
cl::Local(sizeof(float) * work_group_size),
d_partial_sums);
cl::copy(queue, d_partial_sums, h_psum.begin(), h_psum.end());
// complete the sum and compute final integral value
pi_res = 0.0f;
for (unsigned int i = 0; i< nwork_groups; i++) {
pi_res += h_psum[i];
}
pi_res = pi_res * step_size;
//rtime = wtime() - rtime;
double rtime = static_cast<double>(timer.getTimeMilliseconds()) / 1000.;
printf("\nThe calculation ran in %lf seconds\n", rtime);
printf(" pi = %f for %d steps\n", pi_res, nsteps);
}
catch (cl::Error err) {
std::cout << "Exception\n";
std::cerr
<< "ERROR: "
<< err.what()
<< "("
<< err_code(err.err())
<< ")"
<< std::endl;
}
}
/*
* Open an FTDI device. The device is selected with regard to the given restrictions;
* any of those may be NULL and index is counted in the subset selected by the
* description and/or serial if given.
*
* Returns: true if successfully opened, false otherwise.
*/
bool FtdiDevice::open( const char* description, const char* serial, int index )
{
hasFtdiError_ = false;
if ( isOpen() ) {
hasFtdiError_ = false;
return false;
}
context_ = ftdi_new();
if ( context_ == 0 ) {
hasFtdiError_ = true;
return false;
}
bool success = false;
vec_deviceInfo* devs = const_cast<vec_deviceInfo*>( getDeviceList( context_ ) );
vec_deviceInfo::iterator it;
if ( devs != 0 ) {
for ( it = devs->begin(); it != devs->end(); ++it ) {
const deviceInfo& di = *it;
if ( description != 0 || serial != 0 ) {
if ( di.usbInfo == 0 )
break;
if ( description != 0 && std::strncmp( di.usbInfo->description, description,
std::strlen( description ) ) != 0 )
continue;
if ( serial != 0 && std::strncmp( di.usbInfo->serial, serial,
std::strlen( serial ) ) != 0 )
continue;
}
if ( index-- != 0 )
continue;
//If control gets here, we have found a suitable device.
if ( ftdi_usb_open_dev( context_, di.ftdiDevice ) < 0 ) {
//do not free context since that would also free the error message
hasFtdiError_ = true;
}
usbInfo_ = new usbInformation( *( di.usbInfo ) );
success = true;
}
} else {
hasFtdiError_ = ( devs == 0 );
}
if ( ! success && ! hasFtdiError_ ) {
delete context_; context_ = 0;
}
if ( success ) {
purgeBuffers();
reset();
}
return success;
}
static BOOL handleCommand(char line[])
{
BOOL ret = TRUE;
if (strcmp(line, "q") == 0)
ret = FALSE;
else if (strcmp(line, "o") == 0 &&
gDeviceHandle == INVALID_HANDLE_VALUE)
openDriver();
else if (strcmp(line, "c") == 0 &&
gDeviceHandle != INVALID_HANDLE_VALUE)
closeDriver();
else if (strcmp(line, "g") == 0 &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize == 0)
getDeviceList();
else if (strcmp(line, "p") == 0 &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
printDeviceList();
else if (strcmp(line, "r") == 0 &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
releaseDeviceList();
else if (line[0] == 'm' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
sendControlRequest(line + 1);
else if (line[0] == 'a' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
sendAapRequest(line + 1);
else if (line[0] == 'c' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
requestConfigurationDescriptor(line + 1);
else if (line[0] == 'o' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
getActiveConfigValue(line + 1);
else if (line[0] == 's' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
setActiveConfigValue(line + 1);
else if (line[0] == 'i' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
performInterfaceOperation(line + 1);
else if (line[0] == 'r' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
resetDevice(line + 1);
else if (line[0] == 'k' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
performKernelAttachOperation(line + 1);
else if (line[0] == 'b' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
startAAPBulkTransfer(line + 1);
else if (line[0] == 'h' &&
gDeviceHandle != INVALID_HANDLE_VALUE &&
gDeviceListSize > 0)
performHaltOperation(line + 1);
else
printf("Unknown command '%s'\n", line);
return ret;
}
//------------------------------------------------------------
vector<ofSoundDevice> ofSoundStream::listDevices() const{
vector<ofSoundDevice> deviceList = getDeviceList();
ofLogNotice("ofSoundStream::listDevices") << std::endl << deviceList;
return deviceList;
}
int main(int argc, char *argv[])
{
float *h_A; // A matrix
float *h_B; // B matrix
float *h_C; // C = A*B matrix
int N; // A[N][N], B[N][N], C[N][N]
int size; // number of elements in each matrix
cl_mem d_a, d_b, d_c; // Matrices in device memory
double start_time; // Starting time
double run_time; // timing data
char * kernelsource; // kernel source string
cl_int err; // error code returned from OpenCL calls
cl_device_id device; // compute device id
cl_context context; // compute context
cl_command_queue commands; // compute command queue
cl_program program; // compute program
cl_kernel kernel; // compute kernel
N = ORDER;
size = N * N;
h_A = (float *)malloc(size * sizeof(float));
h_B = (float *)malloc(size * sizeof(float));
h_C = (float *)malloc(size * sizeof(float));
//--------------------------------------------------------------------------------
// Create a context, queue and device.
//--------------------------------------------------------------------------------
cl_uint deviceIndex = 0;
parseArguments(argc, argv, &deviceIndex);
// Get list of devices
cl_device_id devices[MAX_DEVICES];
unsigned numDevices = getDeviceList(devices);
// Check device index in range
if (deviceIndex >= numDevices)
{
printf("Invalid device index (try '--list')\n");
return EXIT_FAILURE;
}
device = devices[deviceIndex];
char name[MAX_INFO_STRING];
getDeviceName(device, name);
printf("\nUsing OpenCL device: %s\n", name);
// Create a compute context
context = clCreateContext(0, 1, &device, NULL, NULL, &err);
checkError(err, "Creating context");
// Create a command queue
commands = clCreateCommandQueue(context, device, 0, &err);
checkError(err, "Creating command queue");
//--------------------------------------------------------------------------------
// Run sequential version on the host
//--------------------------------------------------------------------------------
initmat(N, h_A, h_B, h_C);
printf("\n===== Sequential, matrix mult (dot prod), order %d on host CPU ======\n",ORDER);
for(int i = 0; i < COUNT; i++)
{
zero_mat(N, h_C);
start_time = wtime();
seq_mat_mul_sdot(N, h_A, h_B, h_C);
run_time = wtime() - start_time;
results(N, h_C, run_time);
}
//--------------------------------------------------------------------------------
// Setup the buffers, initialize matrices, and write them into global memory
//--------------------------------------------------------------------------------
// Reset A, B and C matrices (just to play it safe)
initmat(N, h_A, h_B, h_C);
d_a = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR,
sizeof(float) * size, h_A, &err);
checkError(err, "Creating buffer d_a");
d_b = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR,
sizeof(float) * size, h_B, &err);
checkError(err, "Creating buffer d_b");
d_c = clCreateBuffer(context, CL_MEM_WRITE_ONLY,
sizeof(float) * size, NULL, &err);
checkError(err, "Creating buffer d_c");
//--------------------------------------------------------------------------------
// OpenCL matrix multiplication ... Naive
//--------------------------------------------------------------------------------
kernelsource = getKernelSource("../C_elem.cl");
// Create the comput program from the source buffer
program = clCreateProgramWithSource(context, 1, (const char **) & kernelsource, NULL, &err);
checkError(err, "Creating program with C_elem.cl");
free(kernelsource);
// Build the program
err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS)
{
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n%s\n", err_code(err));
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
return EXIT_FAILURE;
}
// Create the compute kernel from the program
kernel = clCreateKernel(program, "mmul", &err);
checkError(err, "Creating kernel from C_elem.cl");
printf("\n===== OpenCL, matrix mult, C(i,j) per work item, order %d ======\n",N);
// Do the multiplication COUNT times
for (int i = 0; i < COUNT; i++)
{
zero_mat(N, h_C);
err = clSetKernelArg(kernel, 0, sizeof(int), &N);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 3, sizeof(cl_mem), &d_c);
checkError(err, "Setting kernel args");
start_time = wtime();
// Execute the kernel over the entire range of C matrix elements ... computing
// a dot product for each element of the product matrix. The local work
// group size is set to NULL ... so I'm telling the OpenCL runtime to
// figure out a local work group size for me.
const size_t global[2] = {N, N};
err = clEnqueueNDRangeKernel(
commands,
kernel,
2, NULL,
global, NULL,
0, NULL, NULL);
checkError(err, "Enqueueing kernel");
err = clFinish(commands);
checkError(err, "Waiting for kernel to finish");
run_time = wtime() - start_time;
err = clEnqueueReadBuffer(
commands, d_c, CL_TRUE, 0,
sizeof(float) * size, h_C,
0, NULL, NULL);
checkError(err, "Copying back d_c");
results(N, h_C, run_time);
} // end for loop
//--------------------------------------------------------------------------------
// OpenCL matrix multiplication ... C row per work item
//--------------------------------------------------------------------------------
kernelsource = getKernelSource("../C_row.cl");
// Create the comput program from the source buffer
program = clCreateProgramWithSource(context, 1, (const char **) & kernelsource, NULL, &err);
checkError(err, "Creating program with C_row.cl");
free(kernelsource);
// Build the program
err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS)
{
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n%s\n", err_code(err));
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
return EXIT_FAILURE;
}
// Create the compute kernel from the program
kernel = clCreateKernel(program, "mmul", &err);
checkError(err, "Creating kernel from C_row.cl");
printf("\n===== OpenCL, matrix mult, C row per work item, order %d ======\n",N);
// Do the multiplication COUNT times
for (int i = 0; i < COUNT; i++)
{
zero_mat(N, h_C);
err = clSetKernelArg(kernel, 0, sizeof(int), &N);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 3, sizeof(cl_mem), &d_c);
checkError(err, "Setting kernel args");
start_time = wtime();
// Execute the kernel over the rows of the C matrix ... computing
// a dot product for each element of the product matrix.
const size_t global = N;
err = clEnqueueNDRangeKernel(
commands,
kernel,
1, NULL,
&global, NULL,
0, NULL, NULL);
checkError(err, "Enqueueing kernel");
err = clFinish(commands);
checkError(err, "Waiting for kernel to finish");
run_time = wtime() - start_time;
err = clEnqueueReadBuffer(
commands, d_c, CL_TRUE, 0,
sizeof(float) * size, h_C,
0, NULL, NULL);
checkError(err, "Reading back d_c");
results(N, h_C, run_time);
} // end for loop
//--------------------------------------------------------------------------------
// OpenCL matrix multiplication ... C row per work item, A row in pivate memory
//--------------------------------------------------------------------------------
kernelsource = getKernelSource("../C_row_priv.cl");
// Create the comput program from the source buffer
program = clCreateProgramWithSource(context, 1, (const char **) & kernelsource, NULL, &err);
checkError(err, "Creating program from C_row_priv.cl");
free(kernelsource);
// Build the program
err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS)
{
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n%s\n", err_code(err));
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
return EXIT_FAILURE;
}
// Create the compute kernel from the program
kernel = clCreateKernel(program, "mmul", &err);
checkError(err, "Creating kernel from C_row_priv.cl");
printf("\n===== OpenCL, matrix mult, C row, A row in priv mem, order %d ======\n",N);
// Do the multiplication COUNT times
for (int i = 0; i < COUNT; i++)
{
zero_mat(N, h_C);
err = clSetKernelArg(kernel, 0, sizeof(int), &N);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 3, sizeof(cl_mem), &d_c);
checkError(err, "Setting kernel args");
start_time = wtime();
// Execute the kernel over the rows of the C matrix ... computing
// a dot product for each element of the product matrix.
const size_t global = N;
const size_t local = ORDER / 16;
err = clEnqueueNDRangeKernel(
commands,
kernel,
1, NULL,
&global, &local,
0, NULL, NULL);
checkError(err, "Enqueueing kernel");
err = clFinish(commands);
checkError(err, "Waiting for kernel to finish");
run_time = wtime() - start_time;
err = clEnqueueReadBuffer(
commands, d_c, CL_TRUE, 0,
sizeof(float) * size, h_C,
0, NULL, NULL);
checkError(err, "Reading back d_c");
results(N, h_C, run_time);
} // end for loop
//--------------------------------------------------------------------------------
// Clean up!
//--------------------------------------------------------------------------------
free(h_A);
free(h_B);
free(h_C);
clReleaseMemObject(d_a);
clReleaseMemObject(d_b);
clReleaseMemObject(d_c);
clReleaseProgram(program);
clReleaseKernel(kernel);
clReleaseCommandQueue(commands);
clReleaseContext(context);
return EXIT_SUCCESS;
}
//--------------------------------------------------------------------------------
unsigned int ofxKinectV2::getNumDevices(){
return getDeviceList().size();
}
int main(int argc, char *argv[])
{
float *h_psum; // vector to hold partial sum
int in_nsteps = INSTEPS; // default number of steps (updated later to device preferable)
int niters = ITERS; // number of iterations
int nsteps;
float step_size;
size_t nwork_groups;
size_t max_size, work_group_size = 8;
float pi_res;
cl_mem d_partial_sums;
char *kernelsource = getKernelSource("../pi_ocl.cl"); // Kernel source
cl_int err;
cl_device_id device; // compute device id
cl_context context; // compute context
cl_command_queue commands; // compute command queue
cl_program program; // compute program
cl_kernel kernel_pi; // compute kernel
// Set up OpenCL context, queue, kernel, etc.
cl_uint deviceIndex = 0;
parseArguments(argc, argv, &deviceIndex);
// Get list of devices
cl_device_id devices[MAX_DEVICES];
unsigned numDevices = getDeviceList(devices);
// Check device index in range
if (deviceIndex >= numDevices)
{
printf("Invalid device index (try '--list')\n");
return EXIT_FAILURE;
}
device = devices[deviceIndex];
char name[MAX_INFO_STRING];
getDeviceName(device, name);
printf("\nUsing OpenCL device: %s\n", name);
// Create a compute context
context = clCreateContext(0, 1, &device, NULL, NULL, &err);
checkError(err, "Creating context");
// Create a command queue
commands = clCreateCommandQueue(context, device, 0, &err);
checkError(err, "Creating command queue");
// Create the compute program from the source buffer
program = clCreateProgramWithSource(context, 1, (const char **) & kernelsource, NULL, &err);
checkError(err, "Creating program");
// Build the program
err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS)
{
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n%s\n", err_code(err));
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
return EXIT_FAILURE;
}
// Create the compute kernel from the program
kernel_pi = clCreateKernel(program, "pi", &err);
checkError(err, "Creating kernel");
// Find kernel work-group size
err = clGetKernelWorkGroupInfo (kernel_pi, device, CL_KERNEL_WORK_GROUP_SIZE, sizeof(size_t), &work_group_size, NULL);
checkError(err, "Getting kernel work group info");
// Now that we know the size of the work-groups, we can set the number of
// work-groups, the actual number of steps, and the step size
nwork_groups = in_nsteps/(work_group_size*niters);
if (nwork_groups < 1)
{
err = clGetDeviceInfo(device, CL_DEVICE_MAX_COMPUTE_UNITS, sizeof(size_t), &nwork_groups, NULL);
checkError(err, "Getting device compute unit info");
work_group_size = in_nsteps / (nwork_groups * niters);
}
nsteps = work_group_size * niters * nwork_groups;
step_size = 1.0f/(float)nsteps;
printf("nsteps:%d\n", nsteps);
printf("niters:%d\n", niters);
printf("work_group_size:%zd\n", work_group_size);
printf("n work groups:%ld\n", nwork_groups);
printf("step_size:%f\n", step_size);
h_psum = calloc(sizeof(float), nwork_groups);
if (!h_psum)
{
printf("Error: could not allocate host memory for h_psum\n");
return EXIT_FAILURE;
}
printf(" %ld work-groups of size %ld. %d Integration steps\n",
nwork_groups,
work_group_size,
nsteps);
d_partial_sums = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(float) * nwork_groups, NULL, &err);
checkError(err, "Creating buffer d_partial_sums");
// Set kernel arguments
err = clSetKernelArg(kernel_pi, 0, sizeof(int), &niters);
err |= clSetKernelArg(kernel_pi, 1, sizeof(float), &step_size);
err |= clSetKernelArg(kernel_pi, 2, sizeof(float) * work_group_size, NULL);
err |= clSetKernelArg(kernel_pi, 3, sizeof(cl_mem), &d_partial_sums);
checkError(err, "Settin kernel args");
// Execute the kernel over the entire range of our 1D input data set
// using the maximum number of work items for this device
size_t global = nsteps / niters;
size_t local = work_group_size;
double rtime = wtime();
err = clEnqueueNDRangeKernel(
commands,
kernel_pi,
1, NULL,
&global,
&local,
0, NULL, NULL);
checkError(err, "Enqueueing kernel");
err = clEnqueueReadBuffer(
commands,
d_partial_sums,
CL_TRUE,
0,
sizeof(float) * nwork_groups,
h_psum,
0, NULL, NULL);
checkError(err, "Reading back d_partial_sums");
// complete the sum and compute the final integral value on the host
pi_res = 0.0f;
for (unsigned int i = 0; i < nwork_groups; i++)
{
pi_res += h_psum[i];
}
pi_res *= step_size;
rtime = wtime() - rtime;
printf("\nThe calculation ran in %lf seconds\n", rtime);
printf(" pi = %f for %d steps\n", pi_res, nsteps);
// clean up
clReleaseMemObject(d_partial_sums);
clReleaseProgram(program);
clReleaseKernel(kernel_pi);
clReleaseCommandQueue(commands);
clReleaseContext(context);
free(kernelsource);
free(h_psum);
}
QueryData genDrivers(QueryContext& context) {
QueryData results;
auto devInfoset = setupDevInfoSet();
if (devInfoset == nullptr) {
win32LogWARNING("Error getting device handle");
return results;
}
std::vector<SP_DEVINFO_DATA> devices;
auto ret = getDeviceList(devInfoset, devices);
if (!ret.ok()) {
win32LogWARNING(ret.getMessage(), ret.getCode());
return results;
}
for (auto& device : devices) {
char devId[MAX_DEVICE_ID_LEN] = {0};
if (CM_Get_Device_ID(device.DevInst, devId, MAX_DEVICE_ID_LEN, 0) !=
CR_SUCCESS) {
win32LogWARNING("Failed to get device ID");
return QueryData();
}
SP_DRVINFO_DATA drvInfo = {0};
SP_DRVINFO_DETAIL_DATA drvInfoDetail = {0};
ret = getDeviceDriverInfo(devInfoset, device, drvInfo, drvInfoDetail);
Row r;
r["device_id"] = devId;
r["inf"] = drvInfoDetail.InfFileName;
r["provider"] = drvInfo.ProviderName;
r["manufacturer"] = drvInfo.MfgName;
r["date"] = std::to_string(osquery::filetimeToUnixtime(drvInfo.DriverDate));
r["description"] = drvInfo.Description;
ULARGE_INTEGER version;
version.QuadPart = drvInfo.DriverVersion;
r["version"] = std::to_string(HIWORD(version.HighPart)) + "." +
std::to_string(HIWORD(version.LowPart)) + "." +
std::to_string(LOWORD(version.HighPart)) + "." +
std::to_string(LOWORD(version.LowPart));
for (const auto& elem : kAdditionalDeviceProps) {
std::string val;
ret = getDeviceProperty(devInfoset, device, elem.second, val);
r[elem.first] = std::move(val);
}
if (r.count("driver_key") > 0) {
if (!r.at("driver_key").empty()) {
r["driver_key"].insert(0, kDriverKeyPath);
}
}
if (r.count("service") > 0) {
if (!r.at("service").empty()) {
r["service_key"] = kServiceKeyPath + r["service"];
r["image"] = getDriverImagePath(r["service_key"]);
}
}
results.push_back(r);
}
return results;
}
int main(int argc, char *argv[])
{
int N; // A[N][N], B[N][N], C[N][N]
int size; // Number of elements in each matrix
double start_time; // Starting time
double run_time; // Timing
util::Timer timer; // Timing
N = ORDER;
size = N * N;
std::vector<float> h_A(size); // Host memory for Matrix A
std::vector<float> h_B(size); // Host memory for Matrix B
std::vector<float> h_C(size); // Host memory for Matrix C
cl::Buffer d_a, d_b, d_c; // Matrices in device memory
//--------------------------------------------------------------------------------
// Create a context and queue
//--------------------------------------------------------------------------------
try
{
cl_uint deviceIndex = 0;
parseArguments(argc, argv, &deviceIndex);
// Get list of devices
std::vector<cl::Device> devices;
unsigned numDevices = getDeviceList(devices);
// Check device index in range
if (deviceIndex >= numDevices)
{
std::cout << "Invalid device index (try '--list')\n";
return EXIT_FAILURE;
}
cl::Device device = devices[deviceIndex];
std::string name;
getDeviceName(device, name);
std::cout << "\nUsing OpenCL device: " << name << "\n";
std::vector<cl::Device> chosen_device;
chosen_device.push_back(device);
cl::Context context(chosen_device);
cl::CommandQueue queue(context, device);
//--------------------------------------------------------------------------------
// Run sequential matmul
//--------------------------------------------------------------------------------
initmat(N, h_A, h_B, h_C);
timer.reset();
printf("\n===== Sequential, matrix mult (dot prod), order %d on host CPU ======\n",N);
for(int i = 0; i < COUNT; i++)
{
zero_mat(N, h_C);
start_time = static_cast<double>(timer.getTimeMilliseconds()) / 1000.0;
seq_mat_mul_sdot(N, h_A, h_B, h_C);
run_time = static_cast<double>(timer.getTimeMilliseconds()) / 1000.0 - start_time;
results(N, h_C, run_time);
}
//--------------------------------------------------------------------------------
// Setup the buffers, initialize matrices, and write them into global memory
//--------------------------------------------------------------------------------
// Reset A, B and C matrices (just to play it safe)
initmat(N, h_A, h_B, h_C);
d_a = cl::Buffer(context, h_A.begin(), h_A.end(), true);
d_b = cl::Buffer(context, h_B.begin(), h_B.end(), true);
d_c = cl::Buffer(context, CL_MEM_WRITE_ONLY, sizeof(float) * size);
//--------------------------------------------------------------------------------
// OpenCL matrix multiplication ... Naive
//--------------------------------------------------------------------------------
// Create the compute program from the source buffer
cl::Program program(context, kernelsource, true);
// Create the compute kernel from the program
cl::make_kernel<int, cl::Buffer, cl::Buffer, cl::Buffer> naive_mmul(program, "mmul");
printf("\n===== OpenCL, matrix mult, C(i,j) per work item, order %d ======\n",N);
// Do the multiplication COUNT times
for (int i = 0; i < COUNT; i++)
{
zero_mat(N, h_C);
start_time = static_cast<double>(timer.getTimeMilliseconds()) / 1000.0;
// Execute the kernel over the entire range of C matrix elements ... computing
// a dot product for each element of the product matrix. The local work
// group size is set to NULL ... so I'm telling the OpenCL runtime to
// figure out a local work group size for me.
cl::NDRange global(N, N);
naive_mmul(cl::EnqueueArgs(queue, global),
N, d_a, d_b, d_c);
queue.finish();
run_time = static_cast<double>(timer.getTimeMilliseconds()) / 1000.0 - start_time;
cl::copy(queue, d_c, h_C.begin(), h_C.end());
results(N, h_C, run_time);
} // end for loop
} catch (cl::Error err)
{
std::cout << "Exception\n";
std::cerr << "ERROR: "
<< err.what()
<< "("
<< err_code(err.err())
<< ")"
<< std::endl;
}
return EXIT_SUCCESS;
}
int VideoCaptureV4L2::openDevice(int device, int w, int h, VideoCaptureFormat capFormat) {
std::vector<LinuxCaptureDevice> devices = getDeviceList();
if(device >= devices.size()) {
return 0;
}
LinuxCaptureDevice cap = devices[device];
fd = open(cap.path.c_str(), O_RDWR | O_NONBLOCK, 0);
if(fd == -1) {
printf("ERROR: cannot open: %s\n", cap.path.c_str());
fd = 0;
return 0;
}
// query capabilities
struct v4l2_capability caps;
memset(&caps, 0, sizeof(caps));
if (ioctl(fd, VIDIOC_QUERYCAP, &caps) == -1) {
printf("ERROR: cannot detect capabilities of camera.\n");
return 0;
}
// test if we can use this device as capture device
if(!(caps.capabilities & V4L2_CAP_VIDEO_CAPTURE)) {
printf("ERROR: cannot use device for video capture.\n");
return 0;
}
// check the io type
io_method = LINCAP_IO_METHOD_READ;
bool can_io_readwrite = (caps.capabilities & V4L2_CAP_READWRITE) == V4L2_CAP_READWRITE;
bool can_io_stream = (caps.capabilities & V4L2_CAP_STREAMING) == V4L2_CAP_STREAMING;
if(!can_io_readwrite && !can_io_stream) {
printf("ERROR: cannot use read() or memory streaming with this device.\n");
return 0;
}
if(can_io_stream) {
io_method = LINCAP_IO_METHOD_MMAP;
}
if(!isFormatSupported(capFormat, w, h, 1)) {
printf("ERROR: cannot set pixelformat and size: %dx%d\n", w,h);
return 0;
}
width = w;
height = h;
// initialize IO
if(io_method == LINCAP_IO_METHOD_MMAP) {
initMMAP();
}
else if(io_method == LINCAP_IO_METHOD_READ) {
printf("ERROR: this device can use read()/write() io, but we did not yet program this.\n");
::exit(0);
}
else if(io_method == LINCAP_IO_METHOD_USERPTR) {
printf("ERROR: this device can use i/o with userptr, but we need to program it.\n");
::exit(EXIT_FAILURE);
}
is_opened = true;
return 1;
}
