//! Increments statistics for lose event.
//! \param duration Duration of that game play.
void game_over(const std::chrono::duration<long long>& duration)
{
++m_stats[StatTypes::GAME_LOSES];
m_stats[StatTypes::FASTEST_LOSE] = m_stats[StatTypes::FASTEST_LOSE] != 0 ? std::min(duration.count(), m_stats[StatTypes::FASTEST_LOSE]) : duration.count();
m_stats[StatTypes::SLOWEST_LOSE] = std::max(duration.count(), m_stats[StatTypes::SLOWEST_LOSE]);
m_stats[StatTypes::TOTAL_TIME_PLAYED] += duration.count();
}
//! Increments statistics for win event.
//! \param duration Duration of that game play.
void won(const std::chrono::duration<long long>& duration)
{
++m_stats[StatTypes::GAME_WINS];
m_stats[StatTypes::FASTEST_WIN] = m_stats[StatTypes::FASTEST_WIN] != 0 ? std::min(duration.count(), m_stats[StatTypes::FASTEST_WIN]) : duration.count();
m_stats[StatTypes::SLOWEST_WIN] = std::max(duration.count(), m_stats[StatTypes::SLOWEST_WIN]);
m_stats[StatTypes::TOTAL_TIME_PLAYED] += duration.count();
}
void QVirtualClock::timerCallBack( std::chrono::duration<double>& elapsed_sleep_seconds )
{
UInt32 elapsed_sec = static_cast<UInt32>(elapsed_sleep_seconds.count( ));
if (elapsed_sec < 1)
{
return; //this should never happen because boost::sleep guarantees it.
}
if (!m_update_time_invoked)
{
//This can happen if pdu reception is not yet started, or if for some reason pdus have not
// been received from one timerCallBack invocation to another, or if pdus have stopped completely.
//The solution is to keep virtual time alive
UInt32 elapsed_msec_fraction = (static_cast<UInt32>(elapsed_sleep_seconds.count( ) *1000000)) % 1000000;
UInt64 virtual_time_correction = (static_cast<UInt64>(elapsed_sec) << 32 | elapsed_msec_fraction);
m_virtual_time+=virtual_time_correction;
}
for (UInt32 i = 0; i < elapsed_sec; i++)
{
++m_total_elapsed_seconds;
propagateTimePulseEventToRegistered();
}
m_update_time_invoked = false;
}
void bench() {
warmup();
const int itrNum = 50;
int size = 1024 * 1024 * 1024;
char* s1 = (char*) malloc(sizeof (char) * size);
char* s2 = (char*) malloc(sizeof (char) * size);
for (int i = 0; i < size; i++)
s1[i] = ('a' + i % 26);
int blocksizes[] = {8, 16, 128, 512, 1024,
4096, 1024 * 1024, 1024 * 1024 * 4, 1024 * 1024 * 8,
73, 113, 1024 * 1024 + 97};
{
for (int i = 0; i < sizeof(blocksizes)/sizeof(int); i++) {
const auto t1 = std::chrono::high_resolution_clock::now();
int num = size / blocksizes[i];
for (int itr = 0; itr < itrNum; itr++) {
for (int j = 0; j < num; j++) {
memcpy((s2 + blocksizes[i] * j), (s1 + blocksizes[i] * j), blocksizes[i]);
}
}
const auto t2 = std::chrono::high_resolution_clock::now();
const std::chrono::duration<double> td = t2 - t1;
double result = ((double)itrNum*(size/(1024*1024))) / td.count();
printf("glibc,memcpy:\tblockSize=%d, \t\tresult = %10.6f MB/s,\ttime = %10.6f s\n", blocksizes[i], result, td.count());
}
}
{
for (int i = 0; i < sizeof (blocksizes) / sizeof (int); i++) {
const auto t1 = std::chrono::high_resolution_clock::now();
int num = size / blocksizes[i];
for (int itr = 0; itr < itrNum; itr++) {
for (int j = 0; j < num; j++) {
_i_memcpy_256_unaligned((s2 + blocksizes[i] * j), (s1 + blocksizes[i] * j), blocksizes[i]);
}
}
const auto t2 = std::chrono::high_resolution_clock::now();
const std::chrono::duration<double> td = t2 - t1;
double result = ((double) itrNum * (size / (1024 * 1024))) / td.count();
printf("our poc memcpy:\tblockSize=%d, \t\tresult = %10.6f MB/s,\ttime = %10.6f s\n", blocksizes[i], result, td.count());
}
}
}
int main(int argc, char * const argv[]) {
const int blockSize = 64;
const int numInputChannels = 2;
const int numOutputChannels = 2;
const float sampleRate = 22050.0f;
// pass directory and filename of the patch to load
PdContext *context = zg_context_new(numInputChannels, numOutputChannels, blockSize, sampleRate,
callbackFunction, NULL);
// create a graph from a file
PdGraph *graph = zg_context_new_graph_from_file(context, "/Users/mhroth/workspace/ZenGarden/demo/", "echolon_test0.pd");
if (graph == NULL) {
zg_context_delete(context);
return 1;
}
/*
// create a graph manually
PdGraph *graph = zg_context_new_empty_graph(context);
ZGObject *objOsc = zg_graph_add_new_object(graph, "osc~ 440", 0.0f, 0.0f);
ZGObject *objDac = zg_graph_add_new_object(graph, "dac~", 0.0f, 0.0f);
zg_graph_add_connection(graph, objOsc, 0, objDac, 0);
zg_graph_add_connection(graph, objOsc, 0, objDac, 1);
*/
// attach the graph
zg_graph_attach(graph);
float *inputBuffers = (float *) calloc(numInputChannels * blockSize, sizeof(float));
float *outputBuffers = (float *) calloc(numOutputChannels * blockSize, sizeof(float));
const auto start = std::chrono::high_resolution_clock::now();
for (int i = 0; i < NUM_ITERATIONS; i++) {
//while (1) {
zg_context_process(context, inputBuffers, outputBuffers);
}
const auto end = std::chrono::high_resolution_clock::now();
const std::chrono::duration<double> elapsedTimeSeconds = (end - start); // seconds
const double elapsedTime(elapsedTimeSeconds.count() * 1000.0); // milliseconds
printf("Runtime is: %i iterations in %f milliseconds == %f iterations/second.\n", NUM_ITERATIONS,
elapsedTime, ((double) NUM_ITERATIONS)*1000.0/elapsedTime);
const double simulatedTime = ((double)blockSize / (double)sampleRate) * (double)NUM_ITERATIONS * 1000.0; // milliseconds
printf("Runs in realtime: %s (x%.3f)\n", (simulatedTime >= elapsedTime) ? "YES" : "NO", simulatedTime/elapsedTime);
zg_context_delete(context);
free(inputBuffers);
free(outputBuffers);
return 0;
}
void
SinkClient::update(Observable<std::shared_ptr<VideoFrame>>* /*obs*/,
const std::shared_ptr<VideoFrame>& frame_p)
{
auto& f = *frame_p;
#ifdef DEBUG_FPS
auto currentTime = std::chrono::system_clock::now();
const std::chrono::duration<double> seconds = currentTime - lastFrameDebug_;
++frameCount_;
if (seconds.count() > 1) {
std::ostringstream fps;
fps << frameCount_ / seconds.count();
// Send the framerate in smartInfo
Smartools::getInstance().setFrameRate(id_, fps.str());
frameCount_ = 0;
lastFrameDebug_ = currentTime;
}
#endif
#if HAVE_SHM
// Send the resolution in smartInfo
Smartools::getInstance().setResolution(id_, f.width(), f.height());
shm_->renderFrame(f);
#endif
if (target_.pull) {
VideoFrame dst;
VideoScaler scaler;
const int width = f.width();
const int height = f.height();
#if (defined(__ANDROID__) || defined(__APPLE__))
const int format = VIDEO_PIXFMT_RGBA;
#else
const int format = VIDEO_PIXFMT_BGRA;
#endif
const auto bytes = videoFrameSize(format, width, height);
if (bytes > 0) {
if (auto buffer_ptr = target_.pull(bytes)) {
buffer_ptr->format = libav_utils::libav_pixel_format(format);
buffer_ptr->width = width;
buffer_ptr->height = height;
dst.setFromMemory(buffer_ptr->ptr, format, width, height);
scaler_->scale(f, dst);
target_.push(std::move(buffer_ptr));
}
}
}
}
inline bool retry_for(
const std::chrono::duration<Rep, Period>& total_time,
const std::chrono::duration<Rep, Period>& time_step,
std::function<bool (void)> retried_func) {
assert(time_step.count()!=0);
assert(total_time.count() >= time_step.count());
size_t iterations = total_time.count() / time_step.count();
for(size_t p=0; p <iterations; p++) {
if(retried_func()) { return true; }
std::this_thread::sleep_for(time_step);
}
return false;
}
void dynamic_matrix::assemble_mass()
{
M.resize(mesh.active_dofs(), mesh.active_dofs());
std::vector<doublet<int>> doublets;
doublets.reserve(mesh.active_dofs());
for (auto const& submesh : mesh.meshes())
{
for (std::int64_t element = 0; element < submesh.elements(); element++)
{
auto const dof_view = submesh.local_dof_view(element);
for (std::int64_t p{0}; p < dof_view.size(); ++p)
{
for (std::int64_t q{0}; q < dof_view.size(); ++q)
{
doublets.emplace_back(dof_view(p), dof_view(q));
}
}
}
}
M.setFromTriplets(doublets.begin(), doublets.end());
doublets.clear();
auto const start = std::chrono::steady_clock::now();
for (auto const& submesh : mesh.meshes())
{
for (std::int64_t element = 0; element < submesh.elements(); ++element)
{
auto const& [dofs, m] = submesh.consistent_mass(element);
for (std::int64_t a{0}; a < dofs.size(); a++)
{
for (std::int64_t b{0}; b < dofs.size(); b++)
{
M.add_to(dofs(a), dofs(b), m(a, b));
}
}
}
}
auto const end = std::chrono::steady_clock::now();
std::chrono::duration<double> const elapsed_seconds = end - start;
std::cout << std::string(6, ' ') << "Mass assembly took " << elapsed_seconds.count() << "s\n";
}
FutureReturnCode
spin_node_until_future_complete(
rclcpp::executor::Executor & executor, rclcpp::node::Node::SharedPtr node_ptr,
std::shared_future<ResponseT> & future,
std::chrono::duration<int64_t, TimeT> timeout = std::chrono::duration<int64_t, TimeT>(-1))
{
// TODO(wjwwood): does not work recursively right, can't call spin_node_until_future_complete
// inside a callback executed by an executor.
// Check the future before entering the while loop.
// If the future is already complete, don't try to spin.
std::future_status status = future.wait_for(std::chrono::seconds(0));
auto start_time = std::chrono::system_clock::now();
while (status != std::future_status::ready && rclcpp::utilities::ok()) {
executor.spin_node_once(node_ptr, timeout);
if (timeout.count() >= 0) {
if (start_time + timeout < std::chrono::system_clock::now()) {
return TIMEOUT;
}
}
status = future.wait_for(std::chrono::seconds(0));
}
// If the future completed, and we weren't interrupted by ctrl-C, return the response
if (status == std::future_status::ready) {
return FutureReturnCode::SUCCESS;
}
return FutureReturnCode::INTERRUPTED;
}
// This gets called while rendering frame data is loading into GPU
// Good time to update measurements and physics before rendering next frame!
bool App::frameRenderingQueued(const Ogre::FrameEvent& evt)
{
//calculate delay and show
delay = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::system_clock::now() - last_request);
frames_per_second(delay.count());
if (mShutdown) return false;
if(mRift)
{
if ( mRift->update( evt.timeSinceLastFrame ) )
{
mScene->setRiftPose( mRift->getOrientation(), mRift->getPosition() );
} else {
delete mRift;
mRift = NULL;
}
}
//update the standard input devices
mKeyboard->capture();
mMouse->capture();
//call updates on scene elements
mScene->update( evt.timeSinceLastFrame );
//get now time
last_request = std::chrono::system_clock::now();
//exit if key ESCAPE pressed
if(mKeyboard->isKeyDown(OIS::KC_ESCAPE))
return false;
return true;
}
_cv_status wait_for(std::unique_lock<std::mutex>& lock, const std::chrono::duration<Rep, Period>& rel_time)
{
if (rel_time.count() < 0)
{
return std::cv_status::timeout;
}
duration dur = std::chrono::duration_cast<duration>(rel_time);
if (dur.count() == 0 && rel_time.count() != 0)
{
// don't do zero wait if non-zero wait was requested
dur = duration(1);
}
return wait_for_impl(lock, dur);
}
int main(int argc, char* argv[]){
nome_programa = argv[0];
if(argc == 1) uso();
else configura(argc, argv);
/*-------------------
Solovay-Strassen
-------------------*/
if(SS && !MR && !BPSW){
if(!quiet){
std::cout << "Solovay-Strassen -- Contagem de Tempo: ";
tempo ? std::cout << "ativada.\n" : std::cout << "desativada.\n";
}
while(std::cin >> n >> k){
tempo_decorrido = std::chrono::steady_clock::duration::zero();
inicio = std::chrono::steady_clock::now();
teste = solovay_strassen(n, k);
fim = std::chrono::steady_clock::now();
if(teste) std::cout << n << " e provavelmente primo.";
else std::cout << n << " e composto.";
if(tempo){
tempo_decorrido = std::chrono::duration_cast<std::chrono::duration<double>>(fim - inicio);
std::cout << " Tempo de execucao: " << tempo_decorrido.count();
}
std::cout << std::endl;
}
}
/*----------------
Miller-Rabin
----------------*/
else if(!SS && MR && !BPSW){
void CocoDeviceWrapper::tick()
{
if(video)
{
glClearColor(0.0, 0.0, 0.0, 1.0);
glClear(GL_COLOR_BUFFER_BIT);
glwrap->SwapBuffers();
return;
}
static size_t count = 0;
static double total = 0.0;
static std::chrono::steady_clock::time_point start = std::chrono::steady_clock::now();
static std::chrono::steady_clock::time_point now, last_tick = std::chrono::steady_clock::now(), last_draw = std::chrono::steady_clock::now();
static std::chrono::duration<double, std::milli> tick_dif = std::chrono::duration<double, std::milli>::zero();
static std::chrono::duration<double, std::milli> draw_dif = std::chrono::duration<double, std::milli>::zero();
// Mark time and calc difference since last call
now = std::chrono::steady_clock::steady_clock::now();
draw_dif = now - last_draw;
tick_dif = now - last_tick;
// If the time ellapsed is grater than 16ms we flush OpenGL trying to achieve 60 fps.
if((draw_dif + tick_dif).count() > 16.0)
{
last_draw = now;
if(glwrap)
{
double td = std::chrono::duration_cast<std::chrono::duration<double, std::milli>>(now - start).count();
#ifdef __XMLHTTPREQUEST_HPP__
XMLHttpRequest::tick();
#endif
engine->run(td);
glwrap->SwapBuffers();
if(count++ == 30)
{
count = 1;
total = draw_dif.count();
} else total += draw_dif.count();
}
now = std::chrono::steady_clock::steady_clock::now();
}
last_tick = now;
}
void CameraCallback(CCamera* cam, const void* buffer, int buffer_length)
{
// check ellapsed time and increment fps as long as the the frame is received withthin the specified interval
t2 = std::chrono::high_resolution_clock::now();
time_span = std::chrono::duration_cast<std::chrono::milliseconds>(t2 - t1);
if(time_span.count() < 10 ){
fps += 1;
}else{
printf("fps: %f\n", fps/time_span.count());
startMilliSecondTimer();
fps = 0;
}
//printf("Ellapsed time: %.3f received %d bytes of data\n", time_span.count, buffer_length);
cv::Mat myuv(HEIGHT + HEIGHT/2, WIDTH, CV_8UC1, (unsigned char*)buffer);
cv::Mat mrgb(HEIGHT, WIDTH, CV_8UC4, dest);
cv::cvtColor(myuv, mrgb, CV_YUV2RGBA_NV21);
//cv::imshow("video", mrgb);
//cv::waitKey(1);
}
double IS::GetMOPS(std::chrono::duration<double> total_time,int niter,int num_keys)
{
double mops = 0.0;
double tt = total_time.count();
if( tt > 0 )
{
mops = (double) niter+num_keys;
mops *= niter / (tt*1000000.0);
}
return mops;
}
int main() {
test::a a;
const auto call = [&a](std::size_t len, int index) noexcept {
for(size_t i = 0; i != len; ++ i) a.ary[index]++;
};
constexpr size_t count = 20'0000'0000ll;
const auto t0 = std::chrono::system_clock::now();
std::thread thread{ call, count, test::len };
std::function<void(std::size_t len, int index)>{ call } (count, 0);
thread.join();
const auto t1 = std::chrono::system_clock::now();
const std::chrono::duration<double> d0 = t1 - t0;
std::printf("time: +%lfs\n", d0.count());
std::getchar();
return 0;
}
void SHMSink::render_frame(VideoFrame& src)
{
VideoFrame dst;
VideoScaler scaler;
const int width = src.getWidth();
const int height = src.getHeight();
const int format = VIDEO_PIXFMT_BGRA;
size_t bytes = dst.getSize(width, height, format);
shm_lock();
if (!resize_area(sizeof(SHMHeader) + bytes)) {
ERROR("Could not resize area");
return;
}
dst.setDestination(shm_area_->data, width, height, format);
scaler.scale(src, dst);
#ifdef DEBUG_FPS
const std::chrono::time_point<std::chrono::system_clock> currentTime = std::chrono::system_clock::now();
const std::chrono::duration<double> seconds = currentTime - lastFrameDebug_;
frameCount_++;
if (seconds.count() > 1) {
DEBUG("%s: FPS %f", shm_name_.c_str(), frameCount_ / seconds.count());
frameCount_ = 0;
lastFrameDebug_ = currentTime;
}
#endif
shm_area_->buffer_size = bytes;
shm_area_->buffer_gen++;
sem_post(&shm_area_->notification);
shm_unlock();
}
int main( int argc, char **argv ) {
if( argc < 2 ) {
cerr << "Usage: " << argv[0] << " [FILE]" << endl;
return 0;
}
ifstream ifs(argv[1], ifstream::in);
if(! ifs.good() ) {
cerr << "Unable to open " << argv[1] << endl;
return 0;
}
const int num_tests = 10;
string line;
string val;
int N = 100000;
while( getline( ifs, line ) ) {
vector<int> incr;
stringstream ss;
ss << line;
while( getline( ss, val, ',' ) ) {
int v = stoi( val );
if( v < 1 ) {
cerr << "Error reading input file" << endl;
return 0;
}
incr.push_back( v );
}
cout << "Using increments: " << line << endl;
total_time = std::chrono::steady_clock::duration::zero();
cout << "N = " << N << endl;
for( int i = 0; i < num_tests; ++i ) {
vector<int> vec( N );
generate( vec.begin(), vec.end(), [&]() {
return ( rand() % 6*N ) - 3*N;
} );
ShellSort::Sort( vec, incr );
}
cout << "Average: " << total_time.count() / num_tests << endl;
cout << "------------------------------------" << endl;
N += 100000;
}
return 0;
}
void
ManualTimeKeeper::adjustCloseTime(
std::chrono::duration<std::int32_t> amount)
{
// Copied from TimeKeeper::adjustCloseTime
using namespace std::chrono;
auto const s = amount.count();
std::lock_guard<std::mutex> lock(mutex_);
// Take large offsets, ignore small offsets,
// push the close time towards our wall time.
if (s > 1)
closeOffset_ += seconds((s + 3) / 4);
else if (s < -1)
closeOffset_ += seconds((s - 3) / 4);
else
closeOffset_ = (closeOffset_ * 3) / 4;
}
bool print(Iterator& out, std::chrono::duration<Rep, Period> d) const {
using namespace printers;
if (adaptive) {
auto cnt = d.count();
if (cnt > 1000000000)
return detail::print_numeric(out, cnt / 1000000000)
&& any.print(out, '.')
&& detail::print_numeric(out, (cnt % 1000000000) / 10000000)
&& any.print(out, 's');
if (cnt > 1000000)
return detail::print_numeric(out, cnt / 1000000) && any.print(out, '.')
&& detail::print_numeric(out, (cnt % 1000000) / 10000)
&& str.print(out, "ms");
if (cnt > 1000)
return detail::print_numeric(out, cnt / 1000) && str.print(out, "us");
return detail::print_numeric(out, cnt) && str.print(out, "ns");
}
auto secs = time::duration_cast<time::double_seconds>(d);
return printers::real.print(out, secs.count()) && any.print(out, 's');
}
void computeStatistics(std::chrono::duration<float, std::milli> sampleDurationMilli)
{
const float sampleDurationSeconds= sampleDurationMilli.count() / 1000.f;
const float N = static_cast<float>(sample_count);
// Compute the mean of the error samples, where "error" = abs(omega_sample)
// If we took the mean of the signed omega samples we'd get a value very
// close to zero since the the gyro at rest over a short period has mean-zero noise
PSMVector3f mean_omega_error= *k_psm_float_vector3_zero;
for (int sample_index = 0; sample_index < sample_count; sample_index++)
{
PSMVector3f error_sample= PSM_Vector3fAbs(&omega_samples[sample_index]);
mean_omega_error= PSM_Vector3fAdd(&mean_omega_error, &error_sample);
}
mean_omega_error= PSM_Vector3fUnsafeScalarDivide(&mean_omega_error, N);
// Compute the variance of the (unsigned) sample error, where "error" = abs(omega_sample)
PSMVector3f var_omega= *k_psm_float_vector3_zero;
for (int sample_index = 0; sample_index < sample_count; sample_index++)
{
PSMVector3f error_sample= PSM_Vector3fAbs(&omega_samples[sample_index]);
PSMVector3f diff_from_mean= PSM_Vector3fSubtract(&error_sample, &mean_omega_error);
PSMVector3f diff_from_mean_sqrd= PSM_Vector3fSquare(&diff_from_mean);
var_omega= PSM_Vector3fAdd(&var_omega, &diff_from_mean);
}
var_omega= PSM_Vector3fUnsafeScalarDivide(&var_omega, N - 1);
// Use the max variance of all three axes (should be close)
variance= PSM_Vector3fMaxValue(&var_omega);
// Compute the max drift rate we got across a three axis
PSMVector3f drift_rate= PSM_Vector3fUnsafeScalarDivide(&drift_rotation, sampleDurationSeconds);
PSMVector3f drift_rate_abs= PSM_Vector3fAbs(&drift_rate);
drift= PSM_Vector3fMaxValue(&drift_rate_abs);
}
std::string short_time_format(std::chrono::duration<long long, std::pico> duration)
{
char buff[16];
unsigned long long millis = std::chrono::duration_cast<std::chrono::milliseconds>(duration).count();
unsigned long long micros = std::chrono::duration_cast<std::chrono::microseconds>(duration).count();
unsigned long long nanoss = std::chrono::duration_cast<std::chrono::nanoseconds >(duration).count();
unsigned long long picos = duration.count();
if (millis >= 100) sprintf(buff, "%4lldms", millis);
else if (millis >= 10) sprintf(buff, "%2.1fms", static_cast<double>(micros) / 1000.0);
else if (millis >= 1) sprintf(buff, "%1.2fms", static_cast<double>(micros) / 1000.0);
else if (micros >= 100) sprintf(buff, "%4lldus", micros);
else if (micros >= 10) sprintf(buff, "%2.1fus", static_cast<double>(nanoss) / 1000.0);
else if (micros >= 1) sprintf(buff, "%1.2fus", static_cast<double>(nanoss) / 1000.0);
else if (nanoss >= 100) sprintf(buff, "%4lldns", nanoss);
else if (nanoss >= 10) sprintf(buff, "%2.1fns", static_cast<double>(picos) / 1000.0);
else if (nanoss >= 1) sprintf(buff, "%1.2fns", static_cast<double>(picos) / 1000.0);
else if (picos ) sprintf(buff, "%4lldps", picos);
else
sprintf(buff, "Error!");
return std::string(buff);
}
void tick(std::chrono::duration<float> d)
{
set_relative_location(
{get_relative_location().x, get_relative_location().y + 5 * d.count()});
}
//!
duration(std::chrono::duration<int64_t, std::nano> cpy)
: duration(cpy.count(), std::nano::den) {}
constexpr duration(std::chrono::duration<Rep, Period> d)
: unit(get_time_unit_from_period<Period>()), count(d.count()) {
static_assert(get_time_unit_from_period<Period>() != time_unit::none,
"only seconds, milliseconds or microseconds allowed");
}
int main(int argc, char* argv[]) {
// 1. parse arguments
if (argc != 4) {
print_help(argv[0]);
return EXIT_FAILURE;
}
const std::string name = argv[1];
const int size = std::atoi(argv[2]);
const int count = std::atoi(argv[3]);
if (!is_name_valid(name)) {
printf("Unknown function name '%s'.\n", name.c_str());
return EXIT_FAILURE;
}
if (size <= 0) {
printf("Size must be greater than 0.\n");
return EXIT_FAILURE;
}
if (count <= 0) {
printf("Iteration count must be greater than 0.\n");
return EXIT_FAILURE;
}
// 2. initialize memory
uint8_t* data = static_cast<uint8_t*>(malloc(size));
if (data == nullptr) {
printf("allocation failed");
return EXIT_FAILURE;
}
for (int i=0; i < size; i++) {
data[i] = i;
}
// 3. run the test
size_t n = 0;
size_t k = count;
std::printf("running %-30s...", name.c_str());
std::fflush(stdout);
const auto t1 = std::chrono::high_resolution_clock::now();
if (name == "lookup-8") {
while (k-- > 0) {
n += popcnt_lookup_8bit(data, size);
}
} else if (name == "lookup-64") {
while (k-- > 0) {
n += popcnt_lookup_64bit(data, size);
}
} else if (name == "bit-parallel") {
while (k-- > 0) {
n += popcnt_parallel_64bit_naive(data, size);
}
} else if (name == "bit-parallel-optimized") {
while (k-- > 0) {
n += popcnt_parallel_64bit_optimized(data, size);
}
} else if (name == "sse-bit-parallel") {
while (k-- > 0) {
n += popcnt_SSE_bit_parallel(data, size);
}
} else if (name == "sse-lookup") {
while (k-- > 0) {
n += popcnt_SSE_lookup(data, size);
}
} else if (name == "cpu") {
while (k-- > 0) {
n += popcnt_cpu_64bit(data, size);
}
} else if (name == "lookup-3") {
while (k-- > 0) {
n += popcnt_lookup3_8bit(data, size);
}
} else if (name == "lookup-4") {
while (k-- > 0) {
n += popcnt_lookup4_8bit(data, size);
}
} else if (name == "lookup-7") {
while (k-- > 0) {
n += popcnt_lookup7_8bit(data, size);
}
} else {
assert(false && "wrong function name handling");
}
const auto t2 = std::chrono::high_resolution_clock::now();
const std::chrono::duration<double> td = t2-t1;
printf("reference result = %lu, time = %10.6f s\n", n, td.count());
free(data);
return EXIT_SUCCESS;
}
std::string TestDuration::minutesString(std::chrono::duration<int32_t, std::ratio<60>> dt)
{
return std::to_string(dt.count());
}
std::string TestDuration::hoursString(std::chrono::duration<int32_t, std::ratio<3600>> dt)
{
return std::to_string(dt.count());
}
constexpr duration(std::chrono::duration<Rep, std::ratio<60,1> > d)
: unit(time_unit::seconds), count(d.count() * 60) { }
std::string TestDuration::nanosString(std::chrono::duration<int32_t, std::nano> dt)
{
return std::to_string(dt.count());
}