Having some performance issues on a new windows install. Going to pull device selection out to the GUI and do further testing

master
MitchellHansen 8 years ago
parent b493d56f7a
commit 2421f7bdce

@ -43,6 +43,54 @@ struct device {
bool cl_gl_sharing = false; bool cl_gl_sharing = false;
}; };
struct device_info {
cl_uint cl_device_address_bits;
cl_bool cl_device_available;
cl_bool cl_device_compiler_available;
cl_bool cl_device_endian_little;
cl_bool cl_device_error_correction_support;
char cl_device_extensions[1024];
cl_ulong cl_device_global_mem_cache_size;
cl_uint cl_device_global_mem_cacheline_size;
cl_ulong cl_device_global_mem_size;
cl_bool cl_device_image_support;
size_t cl_device_image2d_max_height;
size_t cl_device_image2d_max_width;
size_t cl_device_image3d_max_depth;
size_t cl_device_image3d_max_height;
size_t cl_device_image3d_max_width;
cl_ulong cl_device_local_mem_size;
cl_uint cl_device_max_clock_frequency;
cl_uint cl_device_max_compute_units;
cl_uint cl_device_max_constant_args;
cl_ulong cl_device_max_constant_buffer_size;
cl_ulong cl_device_max_mem_alloc_size;
size_t cl_device_max_parameter_size;
cl_uint cl_device_max_read_image_args;
cl_uint cl_device_max_samplers;
size_t cl_device_max_work_group_size;
cl_uint cl_device_max_work_item_dimensions;
size_t cl_device_max_work_item_sizes[3];
cl_uint cl_device_max_write_image_args;
cl_uint cl_device_mem_base_addr_align;
cl_uint cl_device_min_data_type_align_size;
char cl_device_name[128];
cl_platform_id cl_device_platform;
cl_uint cl_device_preferred_vector_width_char;
cl_uint cl_device_preferred_vector_width_short;
cl_uint cl_device_preferred_vector_width_int;
cl_uint cl_device_preferred_vector_width_long;
cl_uint cl_device_preferred_vector_width_float;
cl_uint cl_device_preferred_vector_width_double;
char cl_device_profile[256];
size_t cl_device_profiling_timer_resolution;
cl_device_type cl_device_type;
char cl_device_vendor[128];
cl_uint cl_device_vendor_id;
char cl_device_version[128];
char cl_driver_version[128];
};
struct raycaster_settings { struct raycaster_settings {
}; };
@ -98,6 +146,7 @@ public:
// Modify the viewport matrix // Modify the viewport matrix
void test_edit_viewport(int width, int height, float v_fov, float h_fov); void test_edit_viewport(int width, int height, float v_fov, float h_fov);
void gui();
private: private:
@ -105,6 +154,8 @@ private:
// Also checks for the sharing extension // Also checks for the sharing extension
int acquire_platform_and_device(); int acquire_platform_and_device();
int query_hardware();
// With respect to the individual platforms implementation of sharing // With respect to the individual platforms implementation of sharing
// create a shared cl_gl context // create a shared cl_gl context
int create_shared_context(); int create_shared_context();

@ -40,12 +40,12 @@
#include "imgui/imgui-SFML.h" #include "imgui/imgui-SFML.h"
#include "imgui/imgui.h" #include "imgui/imgui.h"
const int WINDOW_X = 1440; const int WINDOW_X = 1536;
const int WINDOW_Y = 900; const int WINDOW_Y = 1024;
const int WORK_SIZE = WINDOW_X * WINDOW_Y; const int WORK_SIZE = WINDOW_X * WINDOW_Y;
const int MAP_X = 512; const int MAP_X = 256;
const int MAP_Y = 512; const int MAP_Y = 256;
const int MAP_Z = 256; const int MAP_Z = 256;
float elap_time(){ float elap_time(){
@ -97,14 +97,14 @@ int main() {
_map.generate_octree(); _map.generate_octree();
_map.a.print_block(0); _map.a.print_block(0);
_map.test_map(); _map.test_map();
std::cin.get(); //std::cin.get();
return 0; //return 0;
// ============================= // =============================
sf::RenderWindow window(sf::VideoMode(WINDOW_X, WINDOW_Y), "SFML"); sf::RenderWindow window(sf::VideoMode(WINDOW_X, WINDOW_Y), "SFML");
window.setMouseCursorVisible(false); window.setMouseCursorVisible(false);
window.setKeyRepeatEnabled(false); window.setKeyRepeatEnabled(false);
window.setFramerateLimit(60); window.setFramerateLimit(120);
window.setVerticalSyncEnabled(false); window.setVerticalSyncEnabled(false);
ImGui::SFML::Init(window); ImGui::SFML::Init(window);
@ -219,7 +219,18 @@ int main() {
fps.frame(delta_time); fps.frame(delta_time);
fps.draw(); fps.draw();
ImGui::Begin("Camera"); ImGuiWindowFlags window_flags = ImGuiWindowFlags_MenuBar;
bool window_show = true;
ImGui::Begin("Camera", &window_show, window_flags);
if (ImGui::BeginMenuBar())
{
if (ImGui::BeginMenu("Menu"))
{
ImGui::EndMenu();
}
ImGui::EndMenuBar();
}
ImGui::Columns(2); ImGui::Columns(2);
@ -277,16 +288,27 @@ int main() {
handle->set_position(light); handle->set_position(light);
} }
// Menu
if (ImGui::CollapsingHeader("Window options"))
{
if (ImGui::TreeNode("Style"))
{
ImGui::ShowStyleEditor();
ImGui::TreePop();
}
}
//light_pos[0] = static_cast<float>(sin(elapsed_time) * 100.0f + 300.0f); //light_pos[0] = static_cast<float>(sin(elapsed_time) * 100.0f + 300.0f);
//light_pos[1] = static_cast<float>(sin(elapsed_time) * 100.0f + 300.0f); //light_pos[1] = static_cast<float>(sin(elapsed_time) * 100.0f + 300.0f);
//sf::Vector3f light(light_pos[0], light_pos[1], light_pos[2]); //sf::Vector3f light(light_pos[0], light_pos[1], light_pos[2]);
//handle->set_position(light); //handle->set_position(light);
ImGui::End(); ImGui::End();
ImGui::Render(); ImGui::Render();

@ -10,6 +10,8 @@ Hardware_Caster::~Hardware_Caster() {
int Hardware_Caster::init() { int Hardware_Caster::init() {
query_hardware();
// Initialize opencl up to the point where we start assigning buffers // Initialize opencl up to the point where we start assigning buffers
error = acquire_platform_and_device(); error = acquire_platform_and_device();
if(vr_assert(error, "aquire_platform_and_device")) if(vr_assert(error, "aquire_platform_and_device"))
@ -265,6 +267,11 @@ void Hardware_Caster::test_edit_viewport(int width, int height, float v_fov, flo
} }
} }
void Hardware_Caster::gui() {
}
int Hardware_Caster::acquire_platform_and_device() { int Hardware_Caster::acquire_platform_and_device() {
// Get the number of platforms // Get the number of platforms
@ -420,6 +427,103 @@ int Hardware_Caster::acquire_platform_and_device() {
return 1; return 1;
}; };
int Hardware_Caster::query_hardware() {
// 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_info>> 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_info>()));
}
// 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);
// Check to see if we even have OpenCL on this machine
if (deviceIdCount == 0) {
std::cout << "There appears to be no devices, or none at least supporting OpenCL" << std::endl;
return OPENCL_NOT_SUPPORTED;
}
// 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 (vr_assert(error, "clGetDeviceIDs"))
return OPENCL_ERROR;
for (unsigned int q = 0; q < deviceIdCount; q++) {
device_info d;
cl_device_id id = deviceIds[q];
clGetDeviceInfo(id, CL_DEVICE_ADDRESS_BITS, sizeof(cl_uint), &d.cl_device_address_bits, NULL);
clGetDeviceInfo(id, CL_DEVICE_AVAILABLE, sizeof(cl_bool), &d.cl_device_available, NULL);
clGetDeviceInfo(id, CL_DEVICE_COMPILER_AVAILABLE, sizeof(cl_bool), &d.cl_device_compiler_available, NULL);
clGetDeviceInfo(id, CL_DEVICE_ENDIAN_LITTLE, sizeof(cl_bool), &d.cl_device_endian_little, NULL);
clGetDeviceInfo(id, CL_DEVICE_ERROR_CORRECTION_SUPPORT, sizeof(cl_bool), &d.cl_device_error_correction_support, NULL);
clGetDeviceInfo(id, CL_DEVICE_EXTENSIONS, sizeof(char)*1024, &d.cl_device_extensions, NULL);
clGetDeviceInfo(id, CL_DEVICE_GLOBAL_MEM_CACHE_SIZE, sizeof(cl_ulong), &d.cl_device_global_mem_cache_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE, sizeof(cl_uint), &d.cl_device_global_mem_cacheline_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_GLOBAL_MEM_SIZE, sizeof(cl_ulong), &d.cl_device_global_mem_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_IMAGE_SUPPORT, sizeof(cl_bool), &d.cl_device_image_support, NULL);
clGetDeviceInfo(id, CL_DEVICE_IMAGE2D_MAX_HEIGHT, sizeof(size_t), &d.cl_device_image2d_max_height, NULL);
clGetDeviceInfo(id, CL_DEVICE_IMAGE2D_MAX_WIDTH, sizeof(size_t), &d.cl_device_image2d_max_width, NULL);
clGetDeviceInfo(id, CL_DEVICE_IMAGE3D_MAX_DEPTH, sizeof(size_t), &d.cl_device_image3d_max_depth, NULL);
clGetDeviceInfo(id, CL_DEVICE_IMAGE3D_MAX_HEIGHT, sizeof(size_t), &d.cl_device_image3d_max_height, NULL);
clGetDeviceInfo(id, CL_DEVICE_IMAGE3D_MAX_WIDTH, sizeof(size_t), &d.cl_device_image3d_max_width, NULL);
clGetDeviceInfo(id, CL_DEVICE_LOCAL_MEM_SIZE, sizeof(cl_ulong), &d.cl_device_local_mem_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_CLOCK_FREQUENCY, sizeof(size_t), &d.cl_device_max_clock_frequency, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_COMPUTE_UNITS, sizeof(size_t), &d.cl_device_max_compute_units, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_CONSTANT_ARGS, sizeof(size_t), &d.cl_device_max_constant_args, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE, sizeof(cl_ulong), &d.cl_device_max_constant_buffer_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(cl_ulong), &d.cl_device_max_mem_alloc_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_PARAMETER_SIZE, sizeof(size_t), &d.cl_device_max_parameter_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_READ_IMAGE_ARGS, sizeof(cl_uint), &d.cl_device_max_read_image_args, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_SAMPLERS, sizeof(cl_uint), &d.cl_device_max_samplers, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_WORK_GROUP_SIZE, sizeof(cl_ulong), &d.cl_device_max_work_group_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS, sizeof(cl_ulong), &d.cl_device_max_work_item_dimensions, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_WORK_ITEM_SIZES, sizeof(size_t)*3, &d.cl_device_max_work_item_sizes, NULL);
clGetDeviceInfo(id, CL_DEVICE_MAX_WRITE_IMAGE_ARGS, sizeof(cl_uint), &d.cl_device_max_write_image_args, NULL);
clGetDeviceInfo(id, CL_DEVICE_MEM_BASE_ADDR_ALIGN, sizeof(cl_uint), &d.cl_device_mem_base_addr_align, NULL);
clGetDeviceInfo(id, CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE, sizeof(cl_uint), &d.cl_device_min_data_type_align_size, NULL);
clGetDeviceInfo(id, CL_DEVICE_NAME, sizeof(char)*128, &d.cl_device_name, NULL);
clGetDeviceInfo(id, CL_DEVICE_PLATFORM, sizeof(cl_platform_id), &d.cl_device_platform, NULL);
clGetDeviceInfo(id, CL_DEVICE_PREFERRED_GLOBAL_ATOMIC_ALIGNMENT, sizeof(cl_uint), &d.cl_device_preferred_vector_width_char, NULL);
clGetDeviceInfo(id, CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT, sizeof(cl_uint), &d.cl_device_preferred_vector_width_short, NULL);
clGetDeviceInfo(id, CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT, sizeof(cl_uint), &d.cl_device_preferred_vector_width_int, NULL);
clGetDeviceInfo(id, CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG, sizeof(cl_uint), &d.cl_device_preferred_vector_width_long, NULL);
clGetDeviceInfo(id, CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT, sizeof(cl_uint), &d.cl_device_preferred_vector_width_float, NULL);
clGetDeviceInfo(id, CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE, sizeof(cl_uint), &d.cl_device_preferred_vector_width_double, NULL);
clGetDeviceInfo(id, CL_DEVICE_PROFILE, sizeof(char) * 256, &d.cl_device_profile, NULL);
clGetDeviceInfo(id, CL_DEVICE_PROFILING_TIMER_RESOLUTION, sizeof(size_t), &d.cl_device_profiling_timer_resolution, NULL);
clGetDeviceInfo(id, CL_DEVICE_TYPE, sizeof(cl_device_type), &d.cl_device_type, NULL);
clGetDeviceInfo(id, CL_DEVICE_VENDOR, sizeof(char)*128, &d.cl_device_vendor, NULL);
clGetDeviceInfo(id, CL_DEVICE_VENDOR_ID, sizeof(cl_uint), &d.cl_device_vendor_id, NULL);
clGetDeviceInfo(id, CL_DEVICE_VERSION, sizeof(char)*128, &d.cl_device_version, NULL);
clGetDeviceInfo(id, CL_DRIVER_VERSION, sizeof(char)*128, &d.cl_driver_version, NULL);
plt_ids.at(d.cl_device_platform).push_back(d);
}
}
return 1;
}
int Hardware_Caster::create_shared_context() { int Hardware_Caster::create_shared_context() {
// Hurray for standards! // Hurray for standards!
@ -526,7 +630,7 @@ int Hardware_Caster::compile_kernel(std::string kernel_source, bool is_path, std
// Try and build the program // Try and build the program
// "-cl-finite-math-only -cl-fast-relaxed-math -cl-unsafe-math-optimizations" // "-cl-finite-math-only -cl-fast-relaxed-math -cl-unsafe-math-optimizations"
error = clBuildProgram(program, 1, &device_id, NULL, NULL, NULL); error = clBuildProgram(program, 1, &device_id, "-cl-finite-math-only -cl-fast-relaxed-math -cl-unsafe-math-optimizations", NULL, NULL);
// Check to see if it errored out // Check to see if it errored out
if (vr_assert(error, "clBuildProgram")) { if (vr_assert(error, "clBuildProgram")) {

Loading…
Cancel
Save