C++ cl::CommandQueue类代码示例

OStack程序员社区-中国程序员成长平台 › 门户 › 编程› C++›C++教程

原作者: [db:作者] 来自: [db:来源] 收藏邀请

本文整理汇总了C++中cl::CommandQueue类的典型用法代码示例。如果您正苦于以下问题：C++ CommandQueue类的具体用法？C++ CommandQueue怎么用？C++ CommandQueue使用的例子？那么恭喜您, 这里精选的类代码示例或许可以为您提供帮助。

在下文中一共展示了CommandQueue类的20个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于我们的系统推荐出更棒的C++代码示例。

示例1: EnqueueAdvancePathsKernel

void PathOCLRenderThread::EnqueueAdvancePathsKernel(cl::CommandQueue &oclQueue) {
	PathOCLRenderEngine *engine = (PathOCLRenderEngine *)renderEngine;
	const u_int taskCount = engine->taskCount;

	// Micro kernels version
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_RT_NEXT_VERTEX, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_HIT_NOTHING, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_HIT_OBJECT, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_RT_DL, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_DL_ILLUMINATE, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_DL_SAMPLE_BSDF, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_GENERATE_NEXT_VERTEX_RAY, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_SPLAT_SAMPLE, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_NEXT_SAMPLE, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
	oclQueue.enqueueNDRangeKernel(*advancePathsKernel_MK_GENERATE_CAMERA_RAY, cl::NullRange,
			cl::NDRange(taskCount), cl::NDRange(advancePathsWorkGroupSize));
}

开发者ID:yangzhengxing，项目名称:luxrender，代码行数:26，代码来源:pathoclthread.cpp

示例2: update

void MetaBallsApp::update()
{
    std::vector<cl::Memory> acquire( { mClParticleBuf, mClMarchingRenderBuffer, mClMarchingDebugBuffer } );
    mClCommandQueue.enqueueAcquireGLObjects( &acquire );
    updateParticles();
    updateMarching();
    mClCommandQueue.enqueueReleaseGLObjects( &acquire );
}

开发者ID:ryanbartley，项目名称:Cinder-OpenCl，代码行数:8，代码来源:MetaBallsApp.cpp

示例3: runKernelLatency

int clPeak::runKernelLatency(cl::CommandQueue &queue, cl::Program &prog, device_info_t &devInfo)
{
    if(!isKernelLatency)
        return 0;

    cl::Context ctx = queue.getInfo<CL_QUEUE_CONTEXT>();
    cl_uint numItems = (devInfo.maxWGSize) * (devInfo.numCUs) * FETCH_PER_WI;
    cl::NDRange globalSize = (numItems / FETCH_PER_WI);
    cl::NDRange localSize = devInfo.maxWGSize;
    int iters = devInfo.kernelLatencyIters;
    float latency;

    try
    {
        log->print(NEWLINE TAB TAB "Kernel launch latency : ");
        log->xmlOpenTag("kernel_launch_latency");
        log->xmlAppendAttribs("unit", "us");

        cl::Buffer inputBuf = cl::Buffer(ctx, CL_MEM_READ_ONLY, (numItems * sizeof(float)));
        cl::Buffer outputBuf = cl::Buffer(ctx, CL_MEM_WRITE_ONLY, (numItems * sizeof(float)));

        cl::Kernel kernel_v1(prog, "global_bandwidth_v1_local_offset");
        kernel_v1.setArg(0, inputBuf), kernel_v1.setArg(1, outputBuf);

        // Dummy calls
        queue.enqueueNDRangeKernel(kernel_v1, cl::NullRange, globalSize, localSize);
        queue.enqueueNDRangeKernel(kernel_v1, cl::NullRange, globalSize, localSize);
        queue.finish();

        latency = 0;
        for(int i=0; i<iters; i++)
        {
            cl::Event timeEvent;
            queue.enqueueNDRangeKernel(kernel_v1, cl::NullRange, globalSize, localSize, NULL, &timeEvent);
            queue.finish();
            cl_ulong start = timeEvent.getProfilingInfo<CL_PROFILING_COMMAND_QUEUED>() / 1000;
            cl_ulong end = timeEvent.getProfilingInfo<CL_PROFILING_COMMAND_START>() / 1000;
            latency += (float)((int)end - (int)start);
        }
        latency /= iters;

        log->print(latency);    log->print(" us" NEWLINE);
        log->xmlSetContent(latency);
        log->xmlCloseTag();
    }
    catch(cl::Error error)
    {
        log->print(error.err() + NEWLINE);
        log->print(TAB TAB "Tests skipped" NEWLINE);
        return -1;
    }

    return 0;
}

开发者ID:nivertech，项目名称:clpeak，代码行数:54，代码来源:kernel_latency.cpp

示例4: enqueueNewMarker

cl::Event RuntimeMeasurementsManager::enqueueNewMarker(cl::CommandQueue queue) {
    cl::Event event;
#if !defined(CL_VERSION_1_2) || defined(CL_USE_DEPRECATED_OPENCL_1_1_APIS)
    // Use deprecated API
    queue.enqueueMarker(&event);
#else
    queue.enqueueMarkerWithWaitList(NULL, &event)
#endif
    queue.finish();

    return event;
}

开发者ID:philliihd，项目名称:OpenCLUtilityLibrary，代码行数:12，代码来源:RuntimeMeasurementManager.cpp

示例5: copyFromDevice

	void copyFromDevice(cl::CommandQueue &queue)
	{
		if(m_pElts==NULL)
			throw cl::Error(CL_INVALID_MEM_OBJECT, "copyFromDevice - Buffer is not initialised.");
		
		queue.enqueueReadBuffer(m_buffer, CL_TRUE, 0, m_cb, m_pElts);
	}

开发者ID:alcides，项目名称:pcg_cl，代码行数:7，代码来源:opencl_buffer.hpp

示例6: kernel

void kernel(cl::Buffer& devOut, cl::CommandQueue& queue)
{
    static std::once_flag   compileFlag;
    static cl::Program      prog;
    static cl::Kernel       kern;

    std::call_once(compileFlag,
        [queue]() {
        prog = cl::Program(queue.getInfo<CL_QUEUE_CONTEXT>(), fractal_ocl_kernel, true);
            kern = cl::Kernel(prog, "julia");
        });

    //auto juliaOp = cl::make_kernel<Buffer, unsigned, unsigned>(kern);

    static const NDRange local(8, 8);
    NDRange global(local[0] * divup(DIMX, local[0]),
                   local[1] * divup(DIMY, local[1]));

    kern.setArg(0, devOut);
    kern.setArg(1, DIMX);
    kern.setArg(2, DIMY);
    queue.enqueueNDRangeKernel(kern, cl::NullRange, global, local);

    //juliaOp(EnqueueArgs(queue, global, local), devOut, DIMX, DIMY);
}

开发者ID:syurkevi，项目名称:forge，代码行数:25，代码来源:fractal.cpp

示例7: runTestType

bool runTestType(cl::Context context, cl::CommandQueue queue)
{
  cl_uint size = 1024 * 2 + 15;
  
  std::vector<T> input(size);
  
  std::cout << "##Testing scan for " << input.size() << " elements and type " 
	    << magnet::CL::detail::traits<T>::kernel_type();
  
  for(size_t i = 0; i < input.size(); ++i)
    input[i] = i+1;
  
  // create input buffer using pinned memory
  cl::Buffer bufferIn(context, CL_MEM_ALLOC_HOST_PTR |
		      CL_MEM_COPY_HOST_PTR | CL_MEM_READ_WRITE, 
		      sizeof(T) * input.size(), &input[0])
    ;
  
  magnet::CL::scan<T> scanFunctor;
  scanFunctor.build(queue, context);
  
  scanFunctor(bufferIn, bufferIn);
  
  std::vector<T> output(size);
  
  queue.enqueueReadBuffer(bufferIn, CL_TRUE, 0, input.size() *
			  sizeof(T), &output[0]);
  bool failed = !testOutput(input, output);

  std::cout << (failed ? " FAILED" : " PASSED") << std::endl; 
  return failed;
}

开发者ID:BigMacchia，项目名称:DynamO，代码行数:32，代码来源:scan_test.cpp

示例8: runKernel

cl::Event runKernel(const cl::CommandQueue& queue, const cl::Kernel& kernel, const cl::NDRange& globalSize, const cl::NDRange& groupSize, std::vector<cl::Event>& events)
{
	cl::Event event;
	queue.enqueueNDRangeKernel(kernel, cl::NullRange, globalSize, groupSize, &events, &event);
	events.push_back(event);
	return event;
}

开发者ID:Jereq，项目名称:Raytracer，代码行数:7，代码来源:CLHelper.cpp

示例9: runTestType

void runTestType(cl::Context context, cl::CommandQueue queue)
{
  cl_uint size = 2 << 10;

  std::vector<T> input(size);

  std::cout << "##Testing bitonic sort for " << input.size() << " elements and type " 
	    << magnet::CL::detail::traits<T>::kernel_type()
	    << std::endl;
  
  for(size_t i = 0; i < input.size(); ++i)
    input[i] = input.size() - i - 1;
  
  // create input buffer using pinned memory
  cl::Buffer bufferIn(context, CL_MEM_ALLOC_HOST_PTR |
		      CL_MEM_COPY_HOST_PTR | CL_MEM_READ_WRITE, 
		      sizeof(T) * input.size(), &input[0])
    ;
  
  magnet::CL::bitonicSort<T> bitonicSortFunctor;
  bitonicSortFunctor.build(queue, context);
  bitonicSortFunctor(bufferIn);

  std::vector<T> output(size);
 
  queue.enqueueReadBuffer(bufferIn, CL_TRUE, 0, input.size() *
			  sizeof(T), &output[0]);

  if (!testOutput(input, output))
    M_throw() << "Incorrect output for size " 
	      << input.size()
	      << " and type "
	      << magnet::CL::detail::traits<T>::kernel_type();
}

开发者ID:BigMacchia，项目名称:DynamO，代码行数:34，代码来源:bitonicsort_test.cpp

示例10: addkernelarg

inline void OpenCL::addkernelarg(std::size_t i, std::vector<T> const & arg, cl::Kernel & kernel,cl::CommandQueue &quene) const
{
    cl::Buffer buffer(this->context,CL_MEM_READ_WRITE,arg.size()*sizeof(T));
//    std::cout << "enqeue\n";
    quene.enqueueWriteBuffer(buffer,CL_FALSE,0,sizeof(T)*arg.size(),&(arg[0]));
    kernel.setArg(i,buffer);

}

开发者ID:crysxd，项目名称:Parallel-Computing-and-Algorithms-X033537-Project，代码行数:8，代码来源:OpenCL.cpp

示例11: const

inline void OpenCL::addkernelarg(std::size_t i, T const (& arg)[N], cl::Kernel & kernel,cl::CommandQueue &quene) const
{
    cl::Buffer buffer(this->context,CL_MEM_READ_WRITE,N*sizeof(T));
//    std::cout << "enqeue\n";
    quene.enqueueWriteBuffer(buffer,CL_FALSE,0,sizeof(T)*N,&arg);
    kernel.setArg(i,buffer);

}

开发者ID:crysxd，项目名称:Parallel-Computing-and-Algorithms-X033537-Project，代码行数:8，代码来源:OpenCL.cpp

示例12: release

    CL::Event OGLSharedFramebuffer::release(CL::CommandQueue& queue, const CL::Event& evt)
    {
        if (_shared) {
            CL::Event e = queue.enq_GL_release(_cl_buffer->get(),
                                               "release framebuffer", evt);
            return e;
        } else {

            assert(_local);
            CL::Event e = queue.enq_read_buffer(*_cl_buffer, _local, _tex_buffer.get_size(),
                                                "read framebuffer", evt);
            queue.wait_for_events(e);

            _tex_buffer.load(_local);
            return CL::Event();
        }
    }

开发者ID:yzhwang，项目名称:micropolis，代码行数:17，代码来源:Framebuffer.cpp

示例13: read

 void read(const cl::CommandQueue &q, size_t offset, size_t size, T *host,
         bool blocking = false) const
 {
     if (size)
         q.enqueueReadBuffer(
                 buffer, blocking ? CL_TRUE : CL_FALSE,
                 sizeof(T) * offset, sizeof(T) * size, host
                 );
 }

开发者ID:1ibrium，项目名称:vexcl，代码行数:9，代码来源:device_vector.hpp

示例14: L2Norm

real L2Norm(const Buffer3D & in,cl::CommandQueue & q)
{
	cl::Buffer ans (CLContextLoader::getContext(),CL_MEM_READ_WRITE,sizeof(real)*in.width()*in.height()*in.depth());

	CLContextLoader::getRedL2NormKer().setArg(0,in());
	CLContextLoader::getRedL2NormKer().setArg(1,ans());

	q.enqueueNDRangeKernel(CLContextLoader::getRedL2NormKer(),
					cl::NDRange(0),
					cl::NDRange(in.width()*in.height()*in.depth()),
					getBestWorkspaceDim(cl::NDRange(in.width()*in.height()*in.depth())));

	ans = performReduction(ans,CLContextLoader::getRedSumAllKer(),q,in.width()*in.height()*in.depth());

	real res;
	q.enqueueReadBuffer(ans,true,0,sizeof(real),&res);
	return sqrt(res);
}

开发者ID:sbabbi，项目名称:multigrid，代码行数:18，代码来源:auxiliary.cpp

示例15: copyToDeviceAsync

	cl::Event copyToDeviceAsync(cl::CommandQueue &queue)
	{
		if(m_pElts==NULL)
			throw cl::Error(CL_INVALID_MEM_OBJECT, "copyToDevice - Buffer is not initialised.");
		
		cl::Event complete;
		queue.enqueueWriteBuffer(m_buffer, CL_FALSE, 0, m_cb, m_pElts, NULL, &complete);
		return complete;
	}

开发者ID:alcides，项目名称:pcg_cl，代码行数:9，代码来源:opencl_buffer.hpp

示例16: acquire

 CL::Event OGLSharedFramebuffer::acquire(CL::CommandQueue& queue, const CL::Event& e)
 {
     if (_shared) {
         return queue.enq_GL_acquire(_cl_buffer->get(),
                                     "acquire framebuffer", e);
     } else {
         return e;
     }
 }

开发者ID:yzhwang，项目名称:micropolis，代码行数:9，代码来源:Framebuffer.cpp

示例17: sumTest

void sumTest(cl::Buffer queue_data, cl::Buffer queue_metadata,
             cl::Buffer& device_result, int iterations,
             ProgramCache& cache,
             cl::CommandQueue& queue)
{
    cl::Context context = queue.getInfo<CL_QUEUE_CONTEXT>();

    std::vector<std::string> sources;
    sources.push_back("ParallelQueue");
    sources.push_back("ParallelQueueTests");

    cl::Program& program = cache.getProgram(sources);

    cl::Kernel sum_test_kernel(program, "sum_test");

    cl::Device device = queue.getInfo<CL_QUEUE_DEVICE>();

    int warp_size = sum_test_kernel
        .getWorkGroupInfo<CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE>(device);

    std::cout << "warp size: " << warp_size << std::endl;

    int max_group_size = device.getInfo<CL_DEVICE_MAX_WORK_ITEM_SIZES>()[0];
    int queue_num_threads = 512;

    if(queue_num_threads > max_group_size)
        queue_num_threads = max_group_size;

    cl::LocalSpaceArg local_queue
            = cl::__local(sizeof(int) * queue_num_threads * 2);
    cl::LocalSpaceArg reduction_buffer
            = cl::__local(sizeof(int) * queue_num_threads);
    cl::LocalSpaceArg got_work
            = cl::__local(sizeof(int));
    cl::LocalSpaceArg prefix_sum_input
            = cl::__local(sizeof(int) * queue_num_threads);
    cl::LocalSpaceArg prefix_sum_output
            = cl::__local(sizeof(int) * queue_num_threads);

    sum_test_kernel.setArg(0, queue_data);
    sum_test_kernel.setArg(1, queue_metadata);
    sum_test_kernel.setArg(2, device_result);
    sum_test_kernel.setArg(3, iterations);
    sum_test_kernel.setArg(4, local_queue);
    sum_test_kernel.setArg(5, reduction_buffer);
    sum_test_kernel.setArg(6, got_work);
    sum_test_kernel.setArg(7, prefix_sum_input);
    sum_test_kernel.setArg(8, prefix_sum_output);

    cl::NDRange nullRange;
    cl::NDRange global(queue_num_threads, 1);
    cl::NDRange local(queue_num_threads, 1);

    cl_int status = queue.enqueueNDRangeKernel(sum_test_kernel,
                                               nullRange, global, local);
}

开发者ID:sharmaashish，项目名称:emory-cci-fast-ia-gsoc，代码行数:56，代码来源:parallelQueueTest.cpp

示例18: clear

 CL::Event Framebuffer::clear(CL::CommandQueue& queue, const CL::Event& e)
 {
     _clear_kernel.set_arg(0, _cl_buffer->get());
     vec4 color = config.clear_color();
     _clear_kernel.set_arg(1, vec4(powf(color.x, 2.2), 
                                   powf(color.y, 2.2),
                                   powf(color.z, 2.2), 1000));
     return queue.enq_kernel(_clear_kernel, _size.x * _size.y, 256,
                             "clear framebuffer", e);
 }

开发者ID:yzhwang，项目名称:micropolis，代码行数:10，代码来源:Framebuffer.cpp

示例19: findMinSeamVert

    void findMinSeamVert(cl::Context &ctx,
                         cl::CommandQueue &cmdQueue,
                         cl::Event &event,
                         std::vector<cl::Event> &deps,
                         cl::Buffer &energyMatrix,
                         cl::Buffer &vertMinEnergy,
                         cl::Buffer &vertMinIdx,
                         int width,
                         int height,
                         int pitch,
                         int colsRemoved) {

        cl_int errNum;
        errNum = findMinSeamVertKernel.setArg(0, energyMatrix);
        errNum |= findMinSeamVertKernel.setArg(1, vertMinEnergy);
        errNum |= findMinSeamVertKernel.setArg(2, vertMinIdx);
        errNum |= findMinSeamVertKernel.setArg(3, cl::__local(256 * sizeof(float)));
        errNum |= findMinSeamVertKernel.setArg(4, cl::__local(256 * sizeof(float)));
        errNum |= findMinSeamVertKernel.setArg(5, width);
        errNum |= findMinSeamVertKernel.setArg(6, height);
        errNum |= findMinSeamVertKernel.setArg(7, pitch);
        errNum |= findMinSeamVertKernel.setArg(8, colsRemoved);

        if (errNum != CL_SUCCESS) {
            std::cerr << "Error setting findMinSeamVert arguments." << std::endl;
            exit(-1);
        }

        // This kernel could be written to use more than one work group, but its probably not worth it.

        cl::NDRange offset = cl::NDRange(0);
        cl::NDRange localWorkSize = cl::NDRange(256);
        cl::NDRange globalWorkSize = cl::NDRange(256);

        errNum = cmdQueue.enqueueNDRangeKernel(findMinSeamVertKernel,
                                               offset,
                                               globalWorkSize,
                                               localWorkSize,
                                               &deps,
                                               &event);
        if (errNum != CL_SUCCESS) {
            std::cerr << "Error enqueuing computeSeams kernel for execution." << std::endl;
            exit(-1);
        }

        /** DEBUG **/
        // int deviceResultIdx[1];
        // float deviceResultEnergy[1];

        // mem::read(ctx, cmdQueue, deviceResultIdx, vertMinIdx);
        // mem::read(ctx, cmdQueue, deviceResultEnergy, vertMinEnergy);

        // std::cout << "deviceResultIdx = " << deviceResultIdx[0] << std::endl;
        // std::cout << "deviceResultEnergy = " << deviceResultEnergy[0] << std::endl;
    }

开发者ID:amidvidy，项目名称:seamcarve-opencl，代码行数:55，代码来源:kernel.hpp

示例20: simulationStep

void simulationStep() {
    try {
        // copy
        auto buffer = cl::Buffer(context, CL_MEM_READ_ONLY,
                                 sizeof(unsigned char) * 4 * fieldWidth * fieldHeight,
                                 nullptr, nullptr);
        queue.enqueueWriteBuffer(buffer, CL_TRUE, 0,
                                 sizeof(unsigned char) * 4 * fieldWidth * fieldHeight,
                                 visualizationBufferCPU, NULL, NULL);

        // enque
        stepKernel.setArg(2, buffer);
        cl::NDRange global((size_t) (fieldWidth * fieldHeight));
        queue.enqueueNDRangeKernel(stepKernel, cl::NullRange, global, cl::NullRange);

        // read back
        queue.enqueueReadBuffer(visualizationBufferGPU, CL_TRUE, 0,
                                sizeof(unsigned char) * 4 * fieldWidth * fieldHeight,
                                visualizationBufferCPU, NULL, NULL);

        // finish
        queue.finish();
    } catch (cl::Error err) {
        std::cout << "Error: " << err.what() << "(" << err.err() << ")" << std::endl;
        exit(3);
    }

    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, fieldWidth, fieldHeight, 0, GL_RGBA, GL_UNSIGNED_BYTE,
                 visualizationBufferCPU);
}

开发者ID:steindani，项目名称:gpgpu_gameoflife，代码行数:30，代码来源:GameOfLife.cpp

注：本文中的cl::CommandQueue类示例由纯净天空整理自Github/MSDocs等源码及文档管理平台，相关代码片段筛选自各路编程大神贡献的开源项目，源码版权归原作者所有，传播和使用请参考对应项目的License；未经允许，请勿转载。