gpgpu/MonteCarlo/MonteCarlo.cpp

// MonteCarlo.cpp : Defines the entry point for the console application.

#include <string>
#include <vector>
#include "Common.h"

// OpenCL C API
#include <CL/opencl.h>

// OpenCL C++ API
#include "cl.hpp"

const bool writeOutRandoms = true;
const size_t randomNumbers = 1024;
const size_t threadCount = 512;

void capi()
{
	// Get a platform ID
	cl_platform_id platformID;
	clGetPlatformIDs(1, &platformID, NULL);
	
	// Get a device ID
	cl_device_id deviceID;
	clGetDeviceIDs(platformID, CL_DEVICE_TYPE_GPU, 1, &deviceID, NULL);
	
	// Create a context
	cl_context context;
	cl_context_properties contextProperties[] =
	{ CL_CONTEXT_PLATFORM, (cl_context_properties)platformID, 0 };
	context = clCreateContext(contextProperties, 1, &deviceID, NULL, NULL, NULL);
	
	// Create a command queue
	cl_command_queue queue;
	queue = clCreateCommandQueue(context, deviceID, CL_QUEUE_PROFILING_ENABLE, NULL);
	
	// Create an OpenCL program
	std::string source = FileToString("../kernels/programs.cl");
	const char* csource = source.c_str();
	cl_program program = clCreateProgramWithSource(context, 1, &csource, NULL, NULL);
	cl_int err = clBuildProgram(program, 1, &deviceID, NULL, NULL, NULL);
	if (err != CL_SUCCESS)
	{
		cl_uint logLength;
		clGetProgramBuildInfo(program, deviceID, CL_PROGRAM_BUILD_LOG, 0, NULL, &logLength);
		char* log = new char[logLength];
		clGetProgramBuildInfo(program, deviceID, CL_PROGRAM_BUILD_LOG, logLength, log, 0);
		std::cout << log << std::endl;
		delete[] log;
		exit(-1);
	}

	// Get the kernel handle
	cl_kernel kernel = clCreateKernel(program, "randomLCG", &err);
	if(!CheckCLError(err)) exit(-1);

	// Allocate memory for random numbers and random seeds
	// Every thread receives a private seed, stored in seedBuffer/clSeedBuffer
	// Every thread is supposed to generate randomNumbers random numbers
	std::vector<float> randomBuffer(threadCount * randomNumbers);
	std::vector<float> seedBuffer(threadCount);
	for (int i = 0; i < threadCount; ++i)
	{
		seedBuffer[i] = rand();
	}

	cl_mem randomBufferDev;
	randomBufferDev = clCreateBuffer(context, CL_MEM_READ_WRITE, sizeof(float)* threadCount * randomNumbers, NULL, &err);
	if (!CheckCLError(err)) exit(-1);
	cl_mem seedBufferDev;
	seedBufferDev = clCreateBuffer(context, CL_MEM_READ_WRITE, sizeof(float) * threadCount, NULL, &err);
	if (!CheckCLError(err)) exit(-1);

	clEnqueueWriteBuffer(queue, seedBufferDev, CL_TRUE, 0, sizeof(float) * threadCount, seedBuffer.data(), 0, NULL, NULL);

	// Set the kernel paramateres
	clSetKernelArg(kernel, 0, sizeof(int), &randomNumbers);
	clSetKernelArg(kernel, 1, sizeof(cl_mem), &seedBufferDev);
	clSetKernelArg(kernel, 2, sizeof(cl_mem), &randomBufferDev);
	// Enqueue the kernel: threadCount threads in total, each generating randomNumbers random numbers in [0,1]
	clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &threadCount, NULL, 0, NULL, NULL);
	// Copy the result back to the host
	clEnqueueReadBuffer(queue, randomBufferDev, CL_TRUE, 0, sizeof(float) * threadCount * randomNumbers, randomBuffer.data(), 0, NULL, NULL);

	// Write out the output
	if(writeOutRandoms == true)
	{
		std::ofstream ofs("randoms.txt", std::ofstream::out);
		for (int i = 0; i < threadCount; ++i) {
			for (int j = 0; j < randomNumbers; ++j) {
				ofs << j << " " << randomBuffer[i + j * threadCount] << std::endl;
			}
			ofs << std::endl;
		}
		ofs.close();
	}

	std::cout << "Finished" << std::endl;
}

void cppapi()
{
	cl_int err = CL_SUCCESS;

	// Get a platform ID
	std::vector<cl::Platform> platforms;
	cl::Platform::get(&platforms);
	if (platforms.size() == 0)
	{
		std::cout << "Unable to find suitable platform." << std::endl;
		exit(-1);
	}

	// Create a context
	cl_context_properties properties[] =
	{ CL_CONTEXT_PLATFORM, (cl_context_properties)(platforms[0])(), 0 };
	cl::Context context(CL_DEVICE_TYPE_GPU, properties);

	// Enumerate the devices
	std::vector<cl::Device> devices = context.getInfo<CL_CONTEXT_DEVICES>();

	// Create the command queue
	cl::Event event;
	cl::CommandQueue queue(context, devices[0], 0, &err);

	// Create the OpenCL program
	std::string programSource = FileToString("../kernels/programs.cl");
	cl::Program program = cl::Program(context, programSource);
	err = program.build(devices);
	if (!CheckCLError(err))
	{
		for (size_t devID = 0; devID < devices.size(); ++devID)
		{
			std::cout << "Device: " << devID << std::endl;
			std::cout << "Build Status: " << program.getBuildInfo<CL_PROGRAM_BUILD_STATUS>(devices[devID]) << std::endl;
			std::cout << "Build Options:\t" << program.getBuildInfo<CL_PROGRAM_BUILD_OPTIONS>(devices[devID]) << std::endl;
			std::cout << "Build Log:\t " << program.getBuildInfo<CL_PROGRAM_BUILD_LOG>(devices[devID]) << std::endl;
			std::cout << "--------------------------------------------------" << std::endl;
		}
	}
	

	// Get the kernel handle
	cl::Kernel kernel(program, "randomLCG", &err);
	CheckCLError(err);

	// Allocate memory for random numbers and random seeds
	// Every thread receives a private seed, stored in seedBuffer/clSeedBuffer
	// Every thread is supposed to generate randomNumbers random numbers
	std::vector<float> randomBuffer(threadCount * randomNumbers);
	std::vector<float> seedBuffer(threadCount);
	for (int i = 0; i < threadCount; ++i)
	{
		seedBuffer[i] = rand();
	}
	
	cl::Buffer clSeedBuffer = cl::Buffer(context, CL_MEM_READ_ONLY, sizeof(float) * threadCount, NULL, &err);
	queue.enqueueWriteBuffer(clSeedBuffer, true, 0, sizeof(float) * threadCount, seedBuffer.data());

	cl::Buffer clRandomBuffer = cl::Buffer(context, CL_MEM_WRITE_ONLY, sizeof(float) * threadCount * randomNumbers, NULL, &err);

	// Set the kernel parameters	
	kernel.setArg(0, randomNumbers);
	kernel.setArg(1, clSeedBuffer);
	kernel.setArg(2, clRandomBuffer);

	// Enqueue the kernel: threadCount threads in total, each generating random numbers in [0,1] randomNumbers times
	queue.enqueueNDRangeKernel(kernel,
		cl::NullRange,
		cl::NDRange(threadCount, 1),
		cl::NullRange,
		NULL,
		&event);
	event.wait();

	// Copy result back to host
	queue.enqueueReadBuffer(clRandomBuffer, true, 0, sizeof(float) * threadCount * randomNumbers, randomBuffer.data());

	// Write out the output to file as "index value" rows
	if(writeOutRandoms == true)
	{
		std::ofstream ofs("randoms2.txt", std::ofstream::out);
		for (int i = 0; i < threadCount; ++i) {
			for (int j = 0; j < randomNumbers; ++j) {
				ofs << j << " " << randomBuffer[i + j * threadCount] << std::endl;
			}
		}
		ofs.close();
	}
	
	std::cout << "Finished" << std::endl;
}

int main()
{
	capi();
	cppapi();
	return 0;
}