Function run_example_shared_memory

example/extra/example_temporary_storage.cpp:55–99

// Host function that runs example_shared_memory kernel
template<class T>
void run_example_shared_memory(size_t size)
{
    constexpr unsigned int block_size = 256;
    // Make sure size is a multiple of block_size
    unsigned int grid_size = (size + block_size - 1) / block_size;
    size                   = block_size * grid_size;

    // Generate input on host and copy it to device
    std::vector<T> host_input = get_random_data<T>(size, 0, 1000);
    // Generating expected output for kernel
    std::vector<T> host_expected_output = get_expected_output<T>(host_input, block_size);
    // For reading device output
    std::vector<T> host_output(size);

    // Device memory allocation
    T* device_input;
    T* device_output;
    HIP_CHECK(hipMalloc(&device_input,
                        host_input.size() * sizeof(typename decltype(host_input)::value_type)));
    HIP_CHECK(hipMalloc(&device_output,
                        host_output.size() * sizeof(typename decltype(host_output)::value_type)));

    // Writing input data to device memory
    hip_write_device_memory<T>(device_input, host_input);

    // Launching kernel example_shared_memory
    hipLaunchKernelGGL(HIP_KERNEL_NAME(example_shared_memory<block_size, T>),
                       dim3(grid_size),
                       dim3(block_size),
                       0,
                       0,
                       device_input,
                       device_output);

    // Reading output from device
    hip_read_device_memory<T>(host_output, device_output);

    // Validating output
    OUTPUT_VALIDATION_CHECK(validate_device_output(host_output, host_expected_output));

    HIP_CHECK(hipFree(device_input));
    HIP_CHECK(hipFree(device_output));

    std::cout << "Kernel run_example_shared_memory run was successful!" << std::endl;
}

// Kernel 2 - storage_type for one primitive union'ed with storage_type of other primitive
template<const unsigned int BlockSize, const unsigned int ItemsPerThread, class T>