Large amount done. OpenCL is almost completely abstracted out to the class

from init to kernel compilation. Next will be getting buffer handling, executing, and reading done.
Changed the kernel to a more minimal example.
Added a jump table for error codes and an assert function for error checking.
Added a routine to compare all the available cl devices and platforms, chooses the high clock of either the GPU or CPU.
master
mitchellhansen 8 years ago
parent ce2623f302
commit 5dea2494a0

@ -1,194 +0,0 @@
#include "Clapper.h"
Clapper::Clapper() {
}
Clapper::~Clapper() {
}
void Clapper::init() {
int error = 0;
// ===================================================================== //
// ==== Opencl setup
// Get the number of platforms
cl_uint platform_count = 0;
clGetPlatformIDs(0, nullptr, &platform_count);
// Fetch the platforms
std::vector<cl_platform_id> platformIds(platform_count);
clGetPlatformIDs(platform_count, platformIds.data(), nullptr);
// get the number of devices, fetch them, choose the first one
cl_uint deviceIdCount = 0;
error = clGetDeviceIDs(platformIds[i], CL_DEVICE_TYPE_ALL, 0, nullptr, &deviceIdCount);
std::vector<cl_device_id> deviceIds(deviceIdCount);
for (int q = 0; q < deviceIdCount; q++) {
std::cout << "++++Device " << q << std::endl;
error = clGetDeviceIDs(platformIds[i], CL_DEVICE_TYPE_ALL, deviceIdCount, deviceIds.data(), NULL);
clGetDeviceInfo(deviceIds[q], CL_DEVICE_NAME, 128, platform, NULL);
clGetDeviceInfo(deviceIds[q], CL_DEVICE_VERSION, 128, version, NULL);
std::cout << platform << "\n";
std::cout << version << "\n\n";
}
std::cout << "============================================" << std::endl;
cl_uint deviceIdCount = 0;
std::vector<cl_device_id> deviceIds;
// Try to get a GPU first
error = clGetDeviceIDs(platformIds[1], CL_DEVICE_TYPE_GPU, 0, nullptr,
&deviceIdCount);
if (deviceIdCount == 0) {
std::cout << "couldn't acquire a GPU, falling back to CPU" << std::endl;
error = clGetDeviceIDs(platformIds[1], CL_DEVICE_TYPE_CPU, 0, nullptr, &deviceIdCount);
deviceIds.resize(deviceIdCount);
error = clGetDeviceIDs(platformIds[1], CL_DEVICE_TYPE_CPU, deviceIdCount, deviceIds.data(), NULL);
}
else {
std::cout << "acquired GPU cl target" << std::endl;
deviceIds.resize(deviceIdCount);
clGetDeviceIDs(platformIds[1], CL_DEVICE_TYPE_GPU, deviceIdCount, deviceIds.data(), nullptr);
}
if (error != 0) {
std::cout << "Err: clGetDeviceIDs returned: " << error << std::endl;
return error;
}
// Hurray for standards!
// Setup the context properties to grab the current GL context
#ifdef linux
cl_context_properties context_properties[] = {
CL_GL_CONTEXT_KHR, (cl_context_properties)glXGetCurrentContext(),
CL_GLX_DISPLAY_KHR, (cl_context_properties)glXGetCurrentDisplay(),
CL_CONTEXT_PLATFORM, (cl_context_properties)platform,
0
};
#elif defined _WIN32
//cl_context_properties context_properties[] = {
// CL_CONTEXT_PLATFORM, (cl_context_properties) platformIds[0],
// CL_GL_CONTEXT_KHR, (cl_context_properties) wglGetCurrentContext(),
// CL_WGL_HDC_KHR, (cl_context_properties) wglGetCurrentDC(),
// 0
//};
HGLRC hGLRC = wglGetCurrentContext();
HDC hDC = wglGetCurrentDC();
cl_context_properties context_properties[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)platformIds[1], CL_GL_CONTEXT_KHR, (cl_context_properties)hGLRC, CL_WGL_HDC_KHR, (cl_context_properties)hDC, 0 };
#elif defined TARGET_OS_MAC
CGLContextObj glContext = CGLGetCurrentContext();
CGLShareGroupObj shareGroup = CGLGetShareGroup(glContext);
cl_context_properties context_properties[] = {
CL_CONTEXT_PROPERTY_USE_CGL_SHAREGROUP_APPLE,
(cl_context_properties)shareGroup,
0
};
#endif
// Create our shared context
auto context = clCreateContext(
context_properties,
1,
&deviceIds[0],
nullptr, nullptr,
&error
);
//cl_device_id devices[32];
//size_t size;
//clGetGLContextInfoKHR(context_properties, CL_DEVICES_FOR_GL_CONTEXT_KHR,
// 32 * sizeof(cl_device_id), devices, &size);
if (error != 0) {
std::cout << "Err: clCreateContext returned: " << error << std::endl;
return error;
}
// And the cl command queue
auto commandQueue = clCreateCommandQueue(context, deviceIds[0], 0, &error);
if (error != 0) {
std::cout << "Err: clCreateCommandQueue returned: " << error << std::endl;
return error;
}
// At this point the shared GL/CL context is up and running
}
void Clapper::print_sys_info() {
int error = 0;
// Get the number of platforms
cl_uint plt_cnt = 0;
clGetPlatformIDs(0, nullptr, &plt_cnt);
// Fetch the platforms
std::vector<cl_platform_id> plt_ids(plt_cnt);
clGetPlatformIDs(plt_cnt, plt_ids.data(), nullptr);
// Print out this machines info
std::cout << "============ Hardware info =================" << std::endl;
for (unsigned int i = 0; i < plt_cnt; i++) {
std::cout << "--Platform: " << i << std::endl;
char platform[128];
char version[128];
clGetPlatformInfo(plt_ids[i], CL_PLATFORM_NAME, 128, platform, NULL);
clGetPlatformInfo(plt_ids[i], CL_PLATFORM_VERSION, 128, version, NULL);
std::cout << platform << "\n";
std::cout << version << "\n\n";
// get the number of devices, fetch them, choose the first one
cl_uint deviceIdCount = 0;
error = clGetDeviceIDs(plt_ids[i], CL_DEVICE_TYPE_ALL, 0, nullptr, &deviceIdCount);
std::vector<cl_device_id> deviceIds(deviceIdCount);
for (int q = 0; q < deviceIdCount; q++) {
std::cout << "++++Device " << q << std::endl;
error = clGetDeviceIDs(plt_ids[i], CL_DEVICE_TYPE_ALL, deviceIdCount, deviceIds.data(), NULL);
clGetDeviceInfo(deviceIds[q], CL_DEVICE_NAME, 128, platform, NULL);
clGetDeviceInfo(deviceIds[q], CL_DEVICE_VERSION, 128, version, NULL);
std::cout << platform << "\n";
std::cout << version << "\n\n";
}
}
std::cout << "============================================" << std::endl;
}

@ -1,48 +0,0 @@
#pragma once
#include <GL/glew.h>
#include <vector>
#include <iostream>
#ifdef linux
#include <CL/cl.h>
#include <CL/opencl.h>
#elif defined _WIN32
#include <CL/cl_gl.h>
#include <CL/cl.h>
#include <CL/opencl.h>
#include <windows.h>
#elif defined TARGET_OS_MAC
# include <OpenGL/OpenGL.h>
# include <OpenCL/opencl.h>
#endif
struct platform_tracker {
};
class Clapper {
public:
enum PLATFORM_TYPE { GPU, CPU, ALL };
Clapper();
~Clapper();
void init();
void print_sys_info();
private:
bool initialized = false;
cl_uint platform_count;
std::vector<cl_platform_id> platformIds;
cl_uint deviceIdCount = 0;
std::vector<cl_device_id> deviceIds;
};

@ -0,0 +1,59 @@
#include <vector>
#include <iostream>
#include "util.hpp"
#include <map>
#ifdef linux
#include <CL/cl.h>
#include <CL/opencl.h>
#elif defined _WIN32
#include <CL/cl_gl.h>
#include <CL/cl.h>
#include <CL/opencl.h>
#include <windows.h>
#elif defined TARGET_OS_MAC
# include <OpenGL/OpenGL.h>
# include <OpenCL/opencl.h>
#endif
struct device {
cl_device_id id;
cl_device_type type;
cl_uint clock_frequency;
char version[128];
cl_platform_id platform;
};
class Clapper {
public:
enum PLATFORM_TYPE { GPU, CPU, ALL };
Clapper();
~Clapper();
int acquire_platform_and_device();
int create_shared_context();
int create_command_queue();
int compile_kernel(std::string kernel_source, bool is_path, std::string kernel_name);
cl_device_id getDeviceID();
cl_platform_id getPlatformID();
cl_context getContext();
private:
int error = 0;
bool initialized = false;
bool assert(int error_code, std::string function_name);
cl_platform_id platform_id;
cl_device_id device_id;
cl_context context;
std::map<std::string, cl_kernel> kernel_map;
};

@ -2,6 +2,8 @@
#include <SFML/Graphics.hpp>
#include <iostream>
#include <math.h>
#include <fstream>
#include <sstream>
const double PI = 3.141592653589793238463;
const float PI_F = 3.14159265358979f;
@ -87,4 +89,17 @@ inline float DegreesToRadians(float in) {
inline float RadiansToDegrees(float in) {
return in * 180.0f / PI;
}
}
inline std::string read_file(std::string file_name){
std::ifstream input_file(file_name);
if (!input_file.is_open()){
std::cout << file_name << " could not be opened" << std::endl;
return nullptr;
}
std::stringstream buf;
buf << input_file.rdbuf();
return buf.str();
}

@ -1,32 +1,17 @@
__constant sampler_t sampler =
CLK_NORMALIZED_COORDS_FALSE
| CLK_ADDRESS_CLAMP_TO_EDGE
| CLK_FILTER_NEAREST;
__constant int FILTER_SIZE = 10;
float FilterValue (__constant const float* filterWeights,
const int x, const int y)
__kernel void hello(__global char* string)
{
return filterWeights[(x+FILTER_SIZE) + (y+FILTER_SIZE)*(FILTER_SIZE*2 + 1)];
}
__kernel void Filter (
__read_only image2d_t input,
__constant float* filterWeights,
__write_only image2d_t output)
{
const int2 pos = {get_global_id(0), get_global_id(1)};
float4 sum = (float4)(0.0f);
for(int y = -FILTER_SIZE; y <= FILTER_SIZE; y++) {
for(int x = -FILTER_SIZE; x <= FILTER_SIZE; x++) {
sum += FilterValue(filterWeights, x, y)
* read_imagef(input, sampler, pos + (int2)(x,y));
}
}
write_imagef (output, (int2)(pos.x, pos.y), sum);
}
string[0] = 'H';
string[1] = 'e';
string[2] = 'l';
string[3] = 'l';
string[4] = 'o';
string[5] = ',';
string[6] = ' ';
string[7] = 'W';
string[8] = 'o';
string[9] = 'r';
string[10] = 'l';
string[11] = 'd';
string[12] = '!';
string[13] = '\0';
}

@ -0,0 +1,398 @@
#include "Clapper.h"
Clapper::Clapper() {
}
Clapper::~Clapper() {
}
int Clapper::acquire_platform_and_device(){
// Get the number of platforms
cl_uint plt_cnt = 0;
clGetPlatformIDs(0, nullptr, &plt_cnt);
// Fetch the platforms
std::map<cl_platform_id, std::vector<device>> plt_ids;
// buffer before map init
std::vector<cl_platform_id> plt_buf(plt_cnt);
clGetPlatformIDs(plt_cnt, plt_buf.data(), nullptr);
// Map init
for (auto id: plt_buf){
plt_ids.emplace(std::make_pair(id, std::vector<device>()));
}
// For each platform, populate its devices
for (unsigned int i = 0; i < plt_cnt; i++) {
cl_uint deviceIdCount = 0;
error = clGetDeviceIDs(plt_buf[i], CL_DEVICE_TYPE_ALL, 0, nullptr, &deviceIdCount);
// Get the device ids
std::vector<cl_device_id> deviceIds(deviceIdCount);
error = clGetDeviceIDs(plt_buf[i], CL_DEVICE_TYPE_ALL, deviceIdCount, deviceIds.data(), NULL);
if (assert(error, "clGetDeviceIDs"))
return -1;
for (int q = 0; q < deviceIdCount; q++) {
device d;
d.id = deviceIds[q];
clGetDeviceInfo(d.id, CL_DEVICE_PLATFORM, 128, &d.platform, NULL);
clGetDeviceInfo(d.id, CL_DEVICE_VERSION, 128, &d.version, NULL);
clGetDeviceInfo(d.id, CL_DEVICE_TYPE, 128, &d.type, NULL);
clGetDeviceInfo(d.id, CL_DEVICE_MAX_CLOCK_FREQUENCY, 128, &d.clock_frequency, NULL);
plt_ids.at(d.platform).push_back(d);
}
}
// The devices how now been queried we want to shoot for a gpu with the fastest clock,
// falling back to the cpu with the fastest clock if we weren't able to find one
device current_best_device;
current_best_device.clock_frequency = 0; // Set this to 0 so the first run always selects a new device
for (auto kvp: plt_ids){
for (auto device: kvp.second){
// Gonna just split this up into cases. There are so many devices I cant test with
// that opencl supports. I'm not going to waste my time making a generic implimentation
// Upon success of a condition, set the current best device values
if (device.type == CL_DEVICE_TYPE_GPU && current_best_device.type != CL_DEVICE_TYPE_GPU){
current_best_device = device;
}
else if (device.clock_frequency > current_best_device.clock_frequency){
current_best_device = device;
}
}
}
platform_id = current_best_device.platform;
device_id = current_best_device.id;
return 0;
};
int Clapper::create_shared_context() {
// Hurray for standards!
// Setup the context properties to grab the current GL context
#ifdef linux
cl_context_properties context_properties[] = {
CL_GL_CONTEXT_KHR, (cl_context_properties)glXGetCurrentContext(),
CL_GLX_DISPLAY_KHR, (cl_context_properties)glXGetCurrentDisplay(),
CL_CONTEXT_PLATFORM, (cl_context_properties)platform,
0
};
#elif defined _WIN32
// TODO: Clean this up next time I'm on a windows machine
//cl_context_properties context_properties[] = {
// CL_CONTEXT_PLATFORM, (cl_context_properties) platformIds[0],
// CL_GL_CONTEXT_KHR, (cl_context_properties) wglGetCurrentContext(),
// CL_WGL_HDC_KHR, (cl_context_properties) wglGetCurrentDC(),
// 0
//};
HGLRC hGLRC = wglGetCurrentContext();
HDC hDC = wglGetCurrentDC();
cl_context_properties context_properties[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)platformIds[1], CL_GL_CONTEXT_KHR, (cl_context_properties)hGLRC, CL_WGL_HDC_KHR, (cl_context_properties)hDC, 0 };
#elif defined TARGET_OS_MAC
CGLContextObj glContext = CGLGetCurrentContext();
CGLShareGroupObj shareGroup = CGLGetShareGroup(glContext);
cl_context_properties context_properties[] = {
CL_CONTEXT_PROPERTY_USE_CGL_SHAREGROUP_APPLE,
(cl_context_properties)shareGroup,
0
};
#endif
// Create our shared context
context = clCreateContext(
context_properties,
1,
&device_id,
nullptr, nullptr,
&error
);
if (assert(error, "clCreateContext"))
return -1;
return 0;
}
int Clapper::create_command_queue(){
// / And the cl command queue
auto commandQueue = clCreateCommandQueue(context, device_id, 0, &error);
if (assert(error, "clCreateCommandQueue"))
return -1;
return 0;
}
int Clapper::compile_kernel(std::string kernel_source, bool is_path, std::string kernel_name) {
const char* source;
if (is_path){
//Load in the kernel, and c stringify it
std::string tmp = read_file(kernel_source);
source = tmp.c_str();
} else {
source = kernel_source.c_str();
}
size_t kernel_source_size = strlen(source);
// Load the source into CL's data structure
cl_program program = clCreateProgramWithSource(
context, 1,
&source,
&kernel_source_size, &error
);
if (assert(error, "clCreateProgramWithSource"))
return -1;
// Try and build the program
error = clBuildProgram(program, 1, &device_id, NULL, NULL, NULL);
// Check to see if it errored out
if (assert(error, "clBuildProgram")){
// Get the size of the queued log
size_t log_size;
clGetProgramBuildInfo(program, device_id, CL_PROGRAM_BUILD_LOG, 0, NULL, &log_size);
char *log = new char[log_size];
// Grab the log
clGetProgramBuildInfo(program, device_id, CL_PROGRAM_BUILD_LOG, log_size, log, NULL);
std::cout << log;
return -1;
}
// Done initializing the kernel
cl_kernel kernel = clCreateKernel(program, kernel_name.c_str(), &error);
if (assert(error, "clCreateKernel"))
return -1;
kernel_map.emplace(std::make_pair(kernel_name, kernel));
}
cl_device_id Clapper::getDeviceID(){ return device_id; };
cl_platform_id Clapper::getPlatformID(){ return platform_id; };
cl_context Clapper::getContext(){ return context; };
bool Clapper::assert(int error_code, std::string function_name){
// Just gonna do a little jump table here, just error codes so who cares
std::string err_msg = "Error : ";
switch (error_code) {
case CL_SUCCESS:
return false;
case CL_DEVICE_NOT_FOUND:
err_msg += "CL_DEVICE_NOT_FOUND";
break;
case CL_DEVICE_NOT_AVAILABLE:
err_msg = "CL_DEVICE_NOT_AVAILABLE";
break;
case CL_COMPILER_NOT_AVAILABLE:
err_msg = "CL_COMPILER_NOT_AVAILABLE";
break;
case CL_MEM_OBJECT_ALLOCATION_FAILURE:
err_msg = "CL_MEM_OBJECT_ALLOCATION_FAILURE";
break;
case CL_OUT_OF_RESOURCES:
err_msg = "CL_OUT_OF_RESOURCES";
break;
case CL_OUT_OF_HOST_MEMORY:
err_msg = "CL_OUT_OF_HOST_MEMORY";
break;
case CL_PROFILING_INFO_NOT_AVAILABLE:
err_msg = "CL_PROFILING_INFO_NOT_AVAILABLE";
break;
case CL_MEM_COPY_OVERLAP:
err_msg = "CL_MEM_COPY_OVERLAP";
break;
case CL_IMAGE_FORMAT_MISMATCH:
err_msg = "CL_IMAGE_FORMAT_MISMATCH";
break;
case CL_IMAGE_FORMAT_NOT_SUPPORTED:
err_msg = "CL_IMAGE_FORMAT_NOT_SUPPORTED";
break;
case CL_BUILD_PROGRAM_FAILURE:
err_msg = "CL_BUILD_PROGRAM_FAILURE";
break;
case CL_MAP_FAILURE:
err_msg = "CL_MAP_FAILURE";
break;
case CL_MISALIGNED_SUB_BUFFER_OFFSET:
err_msg = "CL_MISALIGNED_SUB_BUFFER_OFFSET";
break;
case CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST:
err_msg = "CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST";
break;
case CL_COMPILE_PROGRAM_FAILURE:
err_msg = "CL_COMPILE_PROGRAM_FAILURE";
break;
case CL_LINKER_NOT_AVAILABLE:
err_msg = "CL_LINKER_NOT_AVAILABLE";
break;
case CL_LINK_PROGRAM_FAILURE:
err_msg = "CL_LINK_PROGRAM_FAILURE";
break;
case CL_DEVICE_PARTITION_FAILED:
err_msg = "CL_DEVICE_PARTITION_FAILED";
break;
case CL_KERNEL_ARG_INFO_NOT_AVAILABLE:
err_msg = "CL_KERNEL_ARG_INFO_NOT_AVAILABLE";
break;
case CL_INVALID_VALUE:
err_msg = "CL_INVALID_VALUE";
break;
case CL_INVALID_DEVICE_TYPE:
err_msg = "CL_INVALID_DEVICE_TYPE";
break;
case CL_INVALID_PLATFORM:
err_msg = "CL_INVALID_PLATFORM";
break;
case CL_INVALID_DEVICE:
err_msg = "CL_INVALID_DEVICE";
break;
case CL_INVALID_CONTEXT:
err_msg = "CL_INVALID_CONTEXT";
break;
case CL_INVALID_QUEUE_PROPERTIES:
err_msg = "CL_INVALID_QUEUE_PROPERTIES";
break;
case CL_INVALID_COMMAND_QUEUE:
err_msg = "CL_INVALID_COMMAND_QUEUE";
break;
case CL_INVALID_HOST_PTR:
err_msg = "CL_INVALID_HOST_PTR";
break;
case CL_INVALID_MEM_OBJECT:
err_msg = "CL_INVALID_MEM_OBJECT";
break;
case CL_INVALID_IMAGE_FORMAT_DESCRIPTOR:
err_msg = "CL_INVALID_IMAGE_FORMAT_DESCRIPTOR";
break;
case CL_INVALID_IMAGE_SIZE:
err_msg = "CL_INVALID_IMAGE_SIZE";
break;
case CL_INVALID_SAMPLER:
err_msg = "CL_INVALID_SAMPLER";
break;
case CL_INVALID_BINARY:
err_msg = "CL_INVALID_BINARY";
break;
case CL_INVALID_BUILD_OPTIONS:
err_msg = "CL_INVALID_BUILD_OPTIONS";
break;
case CL_INVALID_PROGRAM:
err_msg = "CL_INVALID_PROGRAM";
break;
case CL_INVALID_PROGRAM_EXECUTABLE:
err_msg = "CL_INVALID_PROGRAM_EXECUTABLE";
break;
case CL_INVALID_KERNEL_NAME:
err_msg = "CL_INVALID_KERNEL_NAME";
break;
case CL_INVALID_KERNEL_DEFINITION:
err_msg = "CL_INVALID_KERNEL_DEFINITION";
break;
case CL_INVALID_KERNEL:
err_msg = "CL_INVALID_KERNEL";
break;
case CL_INVALID_ARG_INDEX:
err_msg = "CL_INVALID_ARG_INDEX";
break;
case CL_INVALID_ARG_VALUE:
err_msg = "CL_INVALID_ARG_VALUE";
break;
case CL_INVALID_ARG_SIZE:
err_msg = "CL_INVALID_ARG_SIZE";
break;
case CL_INVALID_KERNEL_ARGS:
err_msg = "CL_INVALID_KERNEL_ARGS";
break;
case CL_INVALID_WORK_DIMENSION:
err_msg = "CL_INVALID_WORK_DIMENSION";
break;
case CL_INVALID_WORK_GROUP_SIZE:
err_msg = "CL_INVALID_WORK_GROUP_SIZE";
break;
case CL_INVALID_WORK_ITEM_SIZE:
err_msg = "CL_INVALID_WORK_ITEM_SIZE";
break;
case CL_INVALID_GLOBAL_OFFSET:
err_msg = "CL_INVALID_GLOBAL_OFFSET";
break;
case CL_INVALID_EVENT_WAIT_LIST:
err_msg = "CL_INVALID_EVENT_WAIT_LIST";
break;
case CL_INVALID_EVENT:
err_msg = "CL_INVALID_EVENT";
break;
case CL_INVALID_OPERATION:
err_msg = "CL_INVALID_OPERATION";
break;
case CL_INVALID_GL_OBJECT:
err_msg = "CL_INVALID_GL_OBJECT";
break;
case CL_INVALID_BUFFER_SIZE:
err_msg = "CL_INVALID_BUFFER_SIZE";
break;
case CL_INVALID_MIP_LEVEL:
err_msg = "CL_INVALID_MIP_LEVEL";
break;
case CL_INVALID_GLOBAL_WORK_SIZE:
err_msg = "CL_INVALID_GLOBAL_WORK_SIZE";
break;
case CL_INVALID_PROPERTY:
err_msg = "CL_INVALID_PROPERTY";
break;
case CL_INVALID_IMAGE_DESCRIPTOR:
err_msg = "CL_INVALID_IMAGE_DESCRIPTOR";
break;
case CL_INVALID_COMPILER_OPTIONS:
err_msg = "CL_INVALID_COMPILER_OPTIONS";
break;
case CL_INVALID_LINKER_OPTIONS:
err_msg = "CL_INVALID_LINKER_OPTIONS";
break;
case CL_INVALID_DEVICE_PARTITION_COUNT:
err_msg = "CL_INVALID_DEVICE_PARTITION_COUNT";
break;
}
std::cout << err_msg << " =at= " << function_name << std::endl;
return true;
}

@ -1,7 +1,7 @@
#pragma once
#include <cstdio>
#include <cstring>
#include <CL/cl_ext.h>
#include <OpenCL/cl_ext.h>
#include <iostream>
#include <vector>
@ -10,7 +10,6 @@
#elif defined _WIN32
#elif defined TARGET_OS_MAC
# include <GL/glew.h>
# include <OpenGL/OpenGL.h>
# include <OpenCL/opencl.h>
#endif

@ -18,6 +18,8 @@
#elif defined TARGET_OS_MAC
# include <OpenGL/OpenGL.h>
# include <OpenCL/opencl.h>
#include <OpenCL/cl_gl_ext.h>
#include <OpenCL/cl_ext.h>
#endif
#include "TestPlatform.cpp"
@ -25,77 +27,20 @@
#include "Curses.h"
#include "util.hpp"
#include "RayCaster.h"
#include "Clapper.h"
const int WINDOW_X = 150;
const int WINDOW_Y = 150;
std::string read_file(std::string file_name){
std::ifstream input_file(file_name);
if (!input_file.is_open()){
std::cout << file_name << " could not be opened" << std::endl;
return nullptr;
}
std::stringstream buf;
buf << input_file.rdbuf();
return buf.str();
}
int main(){
// ====================================================================== //
// ========== Kernel setup & compilation
// Load in the kernel, and c stringify it
std::string kernel_source;
kernel_source = read_file("../kernels/kernel.txt");
const char* kernel_source_c_str = kernel_source.c_str();
size_t kernel_source_size = strlen(kernel_source_c_str);
// Load the source into CL's data structure
cl_program kernel_program = clCreateProgramWithSource(
context, 1,
&kernel_source_c_str,
&kernel_source_size, &error
);
if (error != 0){
std::cout << "Err: clCreateProgramWithSource returned: " << error << std::endl;
return error;
}
// Try and build the program
error = clBuildProgram(kernel_program, 1, &deviceIds[0], NULL, NULL, NULL);
// Check to see if it errored out
if (error == CL_BUILD_PROGRAM_FAILURE){
// Get the size of the queued log
size_t log_size;
clGetProgramBuildInfo(kernel_program, deviceIds[0], CL_PROGRAM_BUILD_LOG, 0, NULL, &log_size);
char *log = new char[log_size];
Clapper c;
c.acquire_platform_and_device();
c.create_shared_context();
// Grab the log
clGetProgramBuildInfo(kernel_program, deviceIds[0], CL_PROGRAM_BUILD_LOG, log_size, log, NULL);
std::cout << "Err: clBuildProgram returned: " << error << std::endl;
std::cout << log << std::endl;
return error;
}
// Done initializing the kernel
cl_kernel finished_kernel = clCreateKernel(kernel_program, "kernel_name", &error);
if (error != 0){
std::cout << "Err: clCreateKernel returned: " << error << std::endl;
return error;
}
c.compile_kernel("../kernels/kernel.txt", true, "hello");
};
@ -124,23 +69,6 @@ sf::Texture window_texture;
// Y: -1.57 is straight up
// Y: 1.57 is straight down
void test_ray_reflection(){
sf::Vector3f r(0.588, -0.78, -0.196);
sf::Vector3f i(0, 0.928, 0.37);
// is this needed? free spin but bounded 0 < z < pi
if (i.z > PI)
i.z -= PI;
else if (i.z < 0)
i.z += PI;
std::cout << AngleBetweenVectors(r, i);
return;
}
int main0() {

Loading…
Cancel
Save