ALPAKA_ACCELERATOR_NAMESPACE::test Namespace Reference

Classes
struct	KernelAddVectorsD
struct	KernelAddVectorsF

Functions
ALPAKA_FN_ACC void	add_vectors_d (Acc1D const &acc, double const *__restrict__ in1, double const *__restrict__ in2, double *__restrict__ out, uint32_t size)
ALPAKA_FN_ACC void	add_vectors_f (Acc1D const &acc, float const *__restrict__ in1, float const *__restrict__ in2, float *__restrict__ out, uint32_t size)
void	opaque_add_vectors_d (const double *in1, const double *in2, double *out, uint32_t size)
void	opaque_add_vectors_f (const float *in1, const float *in2, float *out, uint32_t size)
void	wrapper_add_vectors_d (Queue &queue, const double *__restrict__ in1, const double *__restrict__ in2, double *__restrict__ out, uint32_t size)
void	wrapper_add_vectors_f (Queue &queue, const float *__restrict__ in1, const float *__restrict__ in2, float *__restrict__ out, uint32_t size)

Function Documentation

add_vectors_d()

ALPAKA_FN_ACC void ALPAKA_ACCELERATOR_NAMESPACE::test::add_vectors_d ( Acc1D const &  acc, double const *__restrict__  in1, double const *__restrict__  in2, double *__restrict__  out, uint32_t  size  )

inline

Definition at line 23 of file DeviceAddition.h.

References mps_fire::i, MillePedeFileConverter_cfg::out, and cms::alpakatools::uniform_elements().

Referenced by ALPAKA_ACCELERATOR_NAMESPACE::test::KernelAddVectorsD::operator()().

                                                         {
    for (auto i : cms::alpakatools::uniform_elements(acc, size)) {
      out[i] = in1[i] + in2[i];
    }
  }

add_vectors_f()

ALPAKA_FN_ACC void ALPAKA_ACCELERATOR_NAMESPACE::test::add_vectors_f ( Acc1D const &  acc, float const *__restrict__  in1, float const *__restrict__  in2, float *__restrict__  out, uint32_t  size  )

inline

Definition at line 13 of file DeviceAddition.h.

References mps_fire::i, MillePedeFileConverter_cfg::out, and cms::alpakatools::uniform_elements().

Referenced by ALPAKA_ACCELERATOR_NAMESPACE::HeterogeneousTestAlpakaDevicePlugins::KernelAddVectorsF::operator()(), and ALPAKA_ACCELERATOR_NAMESPACE::test::KernelAddVectorsF::operator()().

                                                         {
    for (auto i : cms::alpakatools::uniform_elements(acc, size)) {
      out[i] = in1[i] + in2[i];
    }
  }

opaque_add_vectors_d()

void ALPAKA_ACCELERATOR_NAMESPACE::test::opaque_add_vectors_d	(	const double *	in1,
		const double *	in2,
		double *	out,
		uint32_t	size
	)

Definition at line 49 of file DeviceAdditionOpaque.cc.

References MillePedeFileConverter_cfg::out, createBeamHaloJobs::queue, SequenceTypes::wait(), and wrapper_add_vectors_d().

                                                                                              {
    // run on the first available devices
    auto const& device = cms::alpakatools::devices<Platform>()[0];
    Queue queue{device};

    // wrap the input and output data in views
    auto in1_h = cms::alpakatools::make_host_view<const double>(in1, size);
    auto in2_h = cms::alpakatools::make_host_view<const double>(in2, size);
    auto out_h = cms::alpakatools::make_host_view<double>(out, size);

    // allocate input and output buffers on the device
    auto in1_d = cms::alpakatools::make_device_buffer<double[]>(queue, size);
    auto in2_d = cms::alpakatools::make_device_buffer<double[]>(queue, size);
    auto out_d = cms::alpakatools::make_device_buffer<double[]>(queue, size);

    // copy the input data to the device
    // FIXME: pass the explicit size of type uint32_t to avoid compilation error
    // The destination view and the extent are required to have compatible index types!
    alpaka::memcpy(queue, in1_d, in1_h, size);
    alpaka::memcpy(queue, in2_d, in2_h, size);

    // fill the output buffer with zeros
    alpaka::memset(queue, out_d, 0);

    // launch the 1-dimensional kernel for vector addition
    test::wrapper_add_vectors_d(queue, in1_d.data(), in2_d.data(), out_d.data(), size);

    // copy the results from the device to the host
    alpaka::memcpy(queue, out_h, out_d);

    // wait for all the operations to complete
    alpaka::wait(queue);

    // the device buffers are freed automatically
  }

opaque_add_vectors_f()

void ALPAKA_ACCELERATOR_NAMESPACE::test::opaque_add_vectors_f	(	const float *	in1,
		const float *	in2,
		float *	out,
		uint32_t	size
	)

Definition at line 13 of file DeviceAdditionOpaque.cc.

References MillePedeFileConverter_cfg::out, createBeamHaloJobs::queue, SequenceTypes::wait(), and wrapper_add_vectors_f().

Referenced by ALPAKA_ACCELERATOR_NAMESPACE::AlpakaTestOpaqueAdditionModule::analyze().

                                                                                           {
    // run on the first available devices
    auto const& device = cms::alpakatools::devices<Platform>()[0];
    Queue queue{device};

    // wrap the input and output data in views
    auto in1_h = cms::alpakatools::make_host_view<const float>(in1, size);
    auto in2_h = cms::alpakatools::make_host_view<const float>(in2, size);
    auto out_h = cms::alpakatools::make_host_view<float>(out, size);

    // allocate input and output buffers on the device
    auto in1_d = cms::alpakatools::make_device_buffer<float[]>(queue, size);
    auto in2_d = cms::alpakatools::make_device_buffer<float[]>(queue, size);
    auto out_d = cms::alpakatools::make_device_buffer<float[]>(queue, size);

    // copy the input data to the device
    // FIXME: pass the explicit size of type uint32_t to avoid compilation error
    // The destination view and the extent are required to have compatible index types!
    alpaka::memcpy(queue, in1_d, in1_h, size);
    alpaka::memcpy(queue, in2_d, in2_h, size);

    // fill the output buffer with zeros
    alpaka::memset(queue, out_d, 0);

    // launch the 1-dimensional kernel for vector addition
    test::wrapper_add_vectors_f(queue, in1_d.data(), in2_d.data(), out_d.data(), size);

    // copy the results from the device to the host
    alpaka::memcpy(queue, out_h, out_d);

    // wait for all the operations to complete
    alpaka::wait(queue);

    // the device buffers are freed automatically
  }

wrapper_add_vectors_d()

void ALPAKA_ACCELERATOR_NAMESPACE::test::wrapper_add_vectors_d	(	Queue &	queue,
		const double *__restrict__	in1,
		const double *__restrict__	in2,
		double *__restrict__	out,
		uint32_t	size
	)

Definition at line 20 of file DeviceAdditionWrapper.dev.cc.

References MillePedeFileConverter_cfg::out, and createBeamHaloJobs::queue.

Referenced by opaque_add_vectors_d().

                                            {
    alpaka::exec<Acc1D>(queue, cms::alpakatools::make_workdiv<Acc1D>(32, 32), KernelAddVectorsD{}, in1, in2, out, size);
  }

wrapper_add_vectors_f()

void ALPAKA_ACCELERATOR_NAMESPACE::test::wrapper_add_vectors_f	(	Queue &	queue,
		const float *__restrict__	in1,
		const float *__restrict__	in2,
		float *__restrict__	out,
		uint32_t	size
	)

Definition at line 12 of file DeviceAdditionWrapper.dev.cc.

References MillePedeFileConverter_cfg::out, and createBeamHaloJobs::queue.

Referenced by ALPAKA_ACCELERATOR_NAMESPACE::AlpakaTestWrapperAdditionModule::analyze(), and opaque_add_vectors_f().

                                            {
    alpaka::exec<Acc1D>(queue, cms::alpakatools::make_workdiv<Acc1D>(32, 32), KernelAddVectorsF{}, in1, in2, out, size);
  }