int main()
{
Stopwatch stopwatch;
HANDLE hTimer = NULL;
// Create an unnamed waitable timer.
hTimer = CreateWaitableTimer(NULL, TRUE, NULL);
if (NULL == hTimer) {
printf("CreateWaitableTimer failed (%d)\n", GetLastError());
return 1;
}
// With Sleep
stopwatch.Start();
DoWait(hTimer, 1000);
printf("Wait 1000ms, Elapsed time : %d\n", stopwatch.ElapsedTime());
// With Sleep
stopwatch.Start();
DoWait(hTimer, 1234);
Sleep(100);
printf("Wait 1234ms with Sleep(100), Elapsed time : %d\n", stopwatch.ElapsedTime());
return 0;
}
//------------------------------------------------------------------------------
static void
display() {
Stopwatch s;
s.Start();
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glViewport(0, 0, g_width, g_height);
g_hud.FillBackground();
double aspect = g_width/(double)g_height;
identity(g_transformData.ModelViewMatrix);
translate(g_transformData.ModelViewMatrix, -g_pan[0], -g_pan[1], -g_dolly);
rotate(g_transformData.ModelViewMatrix, g_rotate[1], 1, 0, 0);
rotate(g_transformData.ModelViewMatrix, g_rotate[0], 0, 1, 0);
rotate(g_transformData.ModelViewMatrix, -90, 1, 0, 0);
translate(g_transformData.ModelViewMatrix,
-g_center[0], -g_center[1], -g_center[2]);
perspective(g_transformData.ProjectionMatrix,
45.0f, (float)aspect, 0.01f, 500.0f);
multMatrix(g_transformData.ModelViewProjectionMatrix,
g_transformData.ModelViewMatrix,
g_transformData.ProjectionMatrix);
glEnable(GL_DEPTH_TEST);
drawStencils();
// draw the control mesh
g_controlMeshDisplay.Draw(g_stencilOutput->BindSrcBuffer(), 3*sizeof(float),
g_transformData.ModelViewProjectionMatrix);
s.Stop();
float drawCpuTime = float(s.GetElapsed() * 1000.0f);
s.Start();
glFinish();
s.Stop();
float drawGpuTime = float(s.GetElapsed() * 1000.0f);
if (g_hud.IsVisible()) {
g_fpsTimer.Stop();
double fps = 1.0/g_fpsTimer.GetElapsed();
g_fpsTimer.Start();
g_hud.DrawString(10, -100, "# stencils : %d", g_nsamplesDrawn);
g_hud.DrawString(10, -80, "EvalStencils : %.3f ms", g_evalTime);
g_hud.DrawString(10, -60, "GPU Draw : %.3f ms", drawGpuTime);
g_hud.DrawString(10, -40, "CPU Draw : %.3f ms", drawCpuTime);
g_hud.DrawString(10, -20, "FPS : %3.1f", fps);
g_hud.Flush();
}
glFinish();
//checkGLErrors("display leave");
}
//------------------------------------------------------------------------------
void
updateGeom() {
int nverts = (int)g_orgPositions.size() / 3;
std::vector<float> vertex;
vertex.reserve(nverts*6);
const float *p = &g_orgPositions[0];
const float *n = &g_normals[0];
float r = sin(g_frame*0.001f) * g_moveScale;
for (int i = 0; i < nverts; ++i) {
float move = 0.05f*cosf(p[0]*20+g_frame*0.01f);
float ct = cos(p[2] * r);
float st = sin(p[2] * r);
g_positions[i*3+0] = p[0]*ct + p[1]*st;
g_positions[i*3+1] = -p[0]*st + p[1]*ct;
g_positions[i*3+2] = p[2];
p += 3;
}
p = &g_positions[0];
for (int i = 0; i < nverts; ++i) {
vertex.push_back(p[0]);
vertex.push_back(p[1]);
vertex.push_back(p[2]);
vertex.push_back(n[0]);
vertex.push_back(n[1]);
vertex.push_back(n[2]);
p += 3;
n += 3;
}
if (!g_vertexBuffer)
g_vertexBuffer = g_osdmesh->InitializeVertexBuffer(6);
g_vertexBuffer->UpdateData(&vertex[0], nverts);
Stopwatch s;
s.Start();
g_osdmesh->Subdivide(g_vertexBuffer, NULL);
s.Stop();
g_cpuTime = float(s.GetElapsed() * 1000.0f);
s.Start();
g_osdmesh->Synchronize();
s.Stop();
g_gpuTime = float(s.GetElapsed() * 1000.0f);
glBindBuffer(GL_ARRAY_BUFFER, g_vertexBuffer->GetGpuBuffer());
glBindBuffer(GL_ARRAY_BUFFER, 0);
}
int main(int argc, char* argv[])
{
LinearEquationSystemFactory factory;
LinearEquationSystemSolver solver;
for (; n <= nLimit; n *= multiplier)
{
float minTime = FLT_MAX;
for (int attempt = 0; attempt < repeatsNumber; ++attempt)
{
LinearEquationSystem* system = factory.Create(n);
NUMBER* solution = new NUMBER[n];
stopwatch.Start();
solver.Solve(system, solution);
stopwatch.Stop();
float elapsedSeconds = stopwatch.GetElapsedSeconds();
if (elapsedSeconds < minTime)
{
minTime = elapsedSeconds;
}
bool result = checker.IsCorrectSolution(system, solution);
printf("Correct: %s \n", result ? "yes" : "no");
delete system;
delete []solution;
}
printf("N = %d, Elapsed seconds: %f\n", n, minTime);
}
}
static void do_parallel_map(Stopwatch& sw) {
// Create a bunch of subcontexts from here and put them in a map
auto all = create();
::ContextMap<size_t> mp;
for (size_t i = N; i < all.size(); i++)
mp.Add(i, all[i]);
// Now the threads that will make progress hard:
auto proceed = std::make_shared<bool>(true);
for (size_t i = N; i--;)
std::thread([proceed, mp] {
while (*proceed)
for (const auto& cur : mp)
;
}).detach();
auto cleanup = MakeAtExit([&] { *proceed = false; });
std::this_thread::sleep_for(std::chrono::milliseconds(100));
// Enumerate the map while iteration is underway
sw.Start();
for (auto& cur : mp)
;
sw.Stop(n);
}
__attribute__((noinline)) void benchNavigatorNoReloc(VNavigator const *se, SOA3D<Precision> const &points,
SOA3D<Precision> const &dirs, NavStatePool const &inpool,
NavStatePool &outpool)
{
Precision *steps = new Precision[points.size()];
Precision *safeties;
if (WithSafety) safeties = new Precision[points.size()];
Stopwatch timer;
size_t hittargetchecksum = 0L;
timer.Start();
for (decltype(points.size()) i = 0; i < points.size(); ++i) {
if (WithSafety) {
steps[i] = se->ComputeStepAndSafety(points[i], dirs[i], vecgeom::kInfLength, *inpool[i], true, safeties[i]);
} else {
steps[i] = se->ComputeStep(points[i], dirs[i], vecgeom::kInfLength, *inpool[i], *outpool[i]);
}
}
timer.Stop();
std::cerr << timer.Elapsed() << "\n";
double accum(0.), saccum(0.);
for (decltype(points.size()) i = 0; i < points.size(); ++i) {
accum += steps[i];
if (WithSafety) {
saccum += safeties[i];
}
if (outpool[i]->Top()) hittargetchecksum += (size_t)outpool[i]->Top()->id();
}
delete[] steps;
std::cerr << "accum " << se->GetName() << " " << accum << " target checksum " << hittargetchecksum << "\n";
if (WithSafety) {
std::cerr << "saccum " << se->GetName() << " " << saccum << "\n";
}
}
static void
updateGeom() {
int nverts = (int)g_orgPositions.size() / 3;
const float *p = &g_orgPositions[0];
float r = sin(g_frame*0.001f) * g_moveScale;
g_positions.resize(nverts*3);
for (int i = 0; i < nverts; ++i) {
//float move = 0.05f*cosf(p[0]*20+g_frame*0.01f);
float ct = cos(p[2] * r);
float st = sin(p[2] * r);
g_positions[i*3+0] = p[0]*ct + p[1]*st;
g_positions[i*3+1] = -p[0]*st + p[1]*ct;
g_positions[i*3+2] = p[2];
p+=3;
}
Stopwatch s;
s.Start();
// update control points
g_stencilOutput->UpdateData(&g_positions[0], 0, nverts);
// Update random points by applying point & tangent stencils
g_stencilOutput->EvalStencils();
s.Stop();
g_evalTime = float(s.GetElapsed() * 1000.0f);
}
int main ( ) {
using namespace Furrovine;
using namespace Furrovine::Audio;
using namespace Furrovine::Pipeline;
//using namespace Furrovine::Graphics;
const static uint32 samplerate = 48000;
const static uint32 bitspersample = 16;
const static uint32 channelcount = 2;
PCMAudioData data = WavAudioLoader( )( "16bit.wav" );
AudioDevice baseaudiodevice( bitspersample, channelcount, samplerate );
RealtimeAudioDevice audiodevice( baseaudiodevice, milliseconds( 10 ) );
audiodevice.AddDSP( PerChannelDelay( audiodevice.PCMDescription( ), { seconds( 0.513 ), seconds( 0.490 ) }, real( 0.25 ) ) );
AudioBuffer buffer( data.Buffer(), 0, 0, 0, 0, 1 );
audiodevice.SubmitBuffer( buffer );
Stopwatch stopwatch;
stopwatch.Start( );
while ( true ) {
double seconds = stopwatch.ElapsedSeconds( );
if ( seconds < 1 )
continue;
//audiodevice.SubmitBuffer( buffer );
stopwatch.Restart( );
}
}
int benchNoCachingNoVector(Vector3D<Precision> const &point, Vector3D<Precision> const &dir,
std::vector<Vector3D<Precision>> const &corners
#ifdef SORTHITBOXES
,
Container_t &hitlist
#endif
)
{
#ifdef INNERTIMER
Stopwatch timer;
timer.Start();
#endif
int vecsize = corners.size() / 2;
int hitcount = 0;
for (auto box = 0; box < vecsize; ++box) {
double distance = BoxImplementation<translation::kIdentity, rotation::kIdentity>::Intersect(
&corners[2 * box], point, dir, 0, vecgeom::kInfLength);
if (distance < vecgeom::kInfLength) {
hitcount++;
#ifdef SORTHITBOXES
hitlist.push_back(BoxIdDistancePair_t(box, distance));
#endif
}
}
#ifdef INNERTIMER
timer.Stop();
std::cerr << "# ORDINARY hitting " << hitcount << "\n";
std::cerr << "# ORDINARY timer " << timer.Elapsed() << "\n";
#endif
return hitcount;
}
//-------------------------------------------------------------------------------------------------
void Benchmark::DoAll(std::vector<string> vExpr, long num)
{
printf("\n\n\n");
PreprocessExpr(vExpr);
char outstr[400], file[400];
time_t t = time(NULL);
#ifdef _DEBUG
sprintf(outstr, "Bench_%s_%%Y%%m%%d_%%H%%M%%S_dbg.txt", GetShortName().c_str());
#else
sprintf(outstr, "Bench_%s_%%Y%%m%%d_%%H%%M%%S_rel.txt", GetShortName().c_str());
#endif
strftime(file, sizeof(file), outstr, localtime(&t));
FILE *pRes = fopen(file, "w");
assert(pRes);
fprintf(pRes, "# Benchmark results\n");
fprintf(pRes, "# Parser: %s\n", GetName().c_str());
fprintf(pRes, "# Evals per expr: %ld\n", num);
fprintf(pRes, "#\"ms per eval [ms]\", \"evals per sec\", \"Result\", \"expr_len\", \"Expression\"\n");
std::string sExpr;
Stopwatch timer;
timer.Start();
for (std::size_t i = 0; i < vExpr.size(); ++i)
{
try
{
DoBenchmark(vExpr[i], num);
fprintf(pRes, "%4.6lf, %4.6lf, %4.6lf, %d, %s, %s\n", m_fTime1, 1000.0/m_fTime1, m_fResult, (int)vExpr[i].length(), vExpr[i].c_str(), m_sInfo.c_str());
printf( "%4.6lf, %4.6lf, %4.6lf, %d, %s, %s\n", m_fTime1, 1000.0/m_fTime1, m_fResult, (int)vExpr[i].length(), vExpr[i].c_str(), m_sInfo.c_str());
}
catch(...)
{
fprintf(pRes, "fail: %s\n", vExpr[i].c_str());
printf( "fail: %s\n", vExpr[i].c_str());
}
fflush(pRes);
}
double dt = timer.Stop() / ((double)vExpr.size() * num);
fprintf(pRes, "# avg. time per expr.: %lf ms; avg. eval per sec: %lf; total bytecode size: %d",
dt,
1000.0 / dt,
m_nTotalBytecodeSize);
printf("\n#\n# avg. time per expr.: %lf ms; avg. eval per sec: %lf; total bytecode size: %d",
dt,
1000.0 / dt,
m_nTotalBytecodeSize);
fclose(pRes);
}
void playback_exact_CSpace_R3()
{
C2A_Model* P = NULL;
C2A_Model* Q = NULL;
readObjFile(P, "../data/models/CupSpoon/Cup.obj");
readObjFile(Q, "../data/models/CupSpoon/Spoon.obj");
P->ComputeRadius();
Q->ComputeRadius();
Collider3D collider(P, Q);
C2A_Model* CSpace;
readObjFile(CSpace, "../data/cupspoon.obj");
std::vector<ContactSpaceSampleData> contactspace_samples;
std::ifstream in("space_test_3d.txt");
asciiReader(in, contactspace_samples);
std::ofstream timing_file("timing_exact_R3.txt");
std::ofstream PD_file("PD_exact_R3.txt");
for(std::size_t i = 0; i < contactspace_samples.size(); ++i)
{
std::cout << i << std::endl;
DataVector q_col(6);
DataVector q(3);
for(std::size_t j = 0; j < 3; ++j)
q[j] = contactspace_samples[i].v[j];
for(std::size_t j = 0; j < 6; ++j)
q_col[j] = contactspace_samples[i].v[j];
boost::timer t;
//aTimer.Reset();
aTimer.Start();
std::pair<DataVector, double> pd_result;
if(!collider.isCollide(q_col))
{
pd_result.second = 0;
}
else
{
pd_result = Minkowski_Cspace_3D::Exact_PD_R3(q, CSpace);
}
aTimer.Stop();
PD_file << pd_result.second << " ";
//timing_file << t.elapsed() << " ";
timing_file << aTimer.GetTime() * 1000 << " ";
timing_file.flush();
PD_file.flush();
}
}
double
OsdMesh::Synchronize() {
Stopwatch s;
s.Start();
{
_dispatcher->Synchronize();
}
s.Stop();
return s.GetElapsed();
}
void Benchmark::Integer8_Test()
{
double Score ( 0 ),
Cur_Time( 0 );
uint32_t loops ( 0 );
Stopwatch watch;
watch.Start();
for( ; Cur_Time < Time_Per_Test; ++loops, watch.Look( Cur_Time ), Status_Done( static_cast< uint32_t >( Status_Mul*Cur_Time )) )
Score += Intern_Integer8_Test();
m_Integer8_Score = static_cast< uint32_t >(Score/loops);
}
void Benchmark::Integer16_Test()
{
double tmp(0);
double Cur_Time(0);
uint32_t loops=0;
Stopwatch watch;
watch.Start();
for( ; Cur_Time < Time_Per_Test; ++loops, watch.Look( Cur_Time ), Status_Done( static_cast< uint32_t >( Status_Mul*Cur_Time )) )
tmp += Intern_Integer16_Test();
m_Integer16_Score = static_cast< uint32_t >(tmp/loops);
}
void Benchmark::Fp128_Test()
{
if( sizeof( long double ) != 16 )
return;
double tmp(0);
double Cur_Time(0);
uint32_t loops=0;
Stopwatch watch;
watch.Start();
for( ; Cur_Time < Time_Per_Test; ++loops, watch.Look( Cur_Time ), Status_Done( static_cast< uint32_t >( Status_Mul*Cur_Time )) )
tmp += Intern_Fp128_Test();
m_Fp128_Score = static_cast< uint32_t >(tmp/loops);
}
void Test1()
{
DIVIDING_LINE_1('-');
constexpr int Count = 2001001;
#if HAS_BOOST
boost::recursive_mutex osMutex;
ostream_t os(std::cout, osMutex);
#else
ostream_t &os = std::cout;
#endif
Stopwatch sw;
sw.Start();
ProducerConsumerContext context(os, Count, Count / 10, ENABLE_LOGGING);
Producer p1(context, 0, Count / 2, 200);
Producer p2(context, Count / 2, Count - Count / 2, 200);
Consumer c1(context, Count / 3, 200);
Consumer c2(context, Count / 3, 200);
Consumer c3(context, Count - Count / 3 - Count / 3, 200);
std::vector<std::thread> threads;
threads.push_back(p1.Create());
threads.push_back(p2.Create());
threads.push_back(c1.Create());
threads.push_back(c2.Create());
threads.push_back(c3.Create());
for (auto &t : threads)
t.join();
sw.Stop();
std::cout << "Ellapsed time: " << sw.GetElapsedMilliseconds() << "ms" << std::endl;
std::cout << "Checking if test is passed...\n";
sw.Restart();
bool ok = context.IsTestPassed();
sw.Stop();
std::cout << "It takes " << sw.GetElapsedMilliseconds() << "ms to figure out if test is passed." << std::endl;
std::cout << "Is test passed? " << std::boolalpha << ok << std::endl;
}
int main()
{
// Stopwatch object used to measure time intervals
Stopwatch sw;
// Record start time
sw.Start();
// Portion of code to be timed
::Sleep(1000); // Just wait for 1,000 ms (1 second)
// Record end time
sw.Stop();
// Print timing results
cout << "Elapsed time: " << sw.ElapsedMilliseconds() << " ms\n";
}
void StorageDeleteFileChunkJob::Execute()
{
Stopwatch sw;
Buffer filename;
Log_Message("Deleting file chunk %U from disk", chunk->GetChunkID());
sw.Start();
filename.Write(chunk->GetFilename());
delete chunk;
chunk = NULL;
StorageFileDeleter::Delete(filename.GetBuffer());
sw.Stop();
Log_Debug("Deleted, elapsed: %U", (uint64_t) sw.Elapsed());
}
void Test0()
{
DIVIDING_LINE_1('-');
#if !PRODUCER_DO_MY_BEST && !CONSUMER_DO_MY_BEST
# error You are in danger of deadlock!
#endif
constexpr int Count = 2000;
#if HAS_BOOST
boost::recursive_mutex osMutex;
ostream_t os(std::cout, osMutex);
#else
ostream_t &os = std::cout;
#endif
Stopwatch sw;
sw.Start();
ProducerConsumerContext context(os, Count, Count / 10, ENABLE_LOGGING);
Producer p1(context, 0, Count / 2, 200);
Producer p2(context, Count / 2, Count - Count / 2, 200);
Consumer c1(context, Count / 3, 200);
Consumer c2(context, Count / 3, 200);
Consumer c3(context, Count - Count / 3 - Count / 3, 200);
std::vector<std::thread> threads;
threads.push_back(p1.Create());
threads.push_back(p2.Create());
threads.push_back(c1.Create());
threads.push_back(c2.Create());
threads.push_back(c3.Create());
for (auto &t : threads)
t.join();
sw.Stop();
std::cout << "Test Done: " << context.IsTestPassed() << ". Ellapsed time: " << sw.GetElapsedMilliseconds() << "ms\n";
}
double
OsdMesh::Subdivide(OsdVertexBuffer *vertex, OsdVertexBuffer *varying) {
_dispatcher->BindVertexBuffer(vertex, varying);
Stopwatch s;
s.Start();
{
_dispatcher->OnKernelLaunch();
_farMesh->Subdivide(_level+1, _exact);
_dispatcher->OnKernelFinish();
}
s.Stop();
_dispatcher->UnbindVertexBuffer();
return s.GetElapsed();
}
void playback_local_PD_R3()
{
std::ofstream timing_file("timing_local_PD_R3.txt");
std::ofstream PD_file("PD_local_R3.txt");
std::vector<C2A_Model*> P;
std::vector<C2A_Model*> Q;
readObjFiles(P, "../data/models/CupSpoon/cup_convex.obj");
readObjFiles(Q, "../data/models/CupSpoon/spoon_convex.obj");
std::vector<ContactSpaceSampleData> contactspace_samples;
std::ifstream in("space_test_3d.txt");
asciiReader(in, contactspace_samples);
for(std::size_t i = 0; i < contactspace_samples.size(); ++i)
{
std::cout << i << std::endl;
DataVector q_col(6);
DataVector q(3);
for(std::size_t j = 0; j < 3; ++j)
q[j] = contactspace_samples[i].v[j];
for(std::size_t j = 0; j < 6; ++j)
q_col[j] = contactspace_samples[i].v[j];
boost::timer t;
aTimer.Reset();
aTimer.Start();
double pd = Collider3D::PDt(P, Q, q_col);
PD_file << pd << " ";
// timing_file << t.elapsed() << " ";
timing_file << aTimer.GetTime() * 1000 << " ";
timing_file.flush();
PD_file.flush();
}
}
static void do_parallel_enum(Stopwatch& sw) {
AutoCurrentContext ctxt;
auto all = create();
// Create threads which will cause contention:
auto proceed = std::make_shared<bool>(true);
for (size_t nParallel = N; nParallel--;) {
std::thread([proceed, all, ctxt] {
while (*proceed)
for (auto cur : ContextEnumerator(ctxt))
;
}).detach();
}
auto cleanup = MakeAtExit([&] { *proceed = false; });
std::this_thread::sleep_for(std::chrono::milliseconds(100));
// Perform parallel enumeration
sw.Start();
for (auto cur : ContextEnumerator(ctxt))
;
sw.Stop(n);
}
Benchmark::Test_Re_Type Benchmark::Memory_DA_Test()
{
Stopwatch watch;
double Comp_Time(0);
if( !::Force_No_Threads )
{
bool Blocker( true );
Smart_Array_Pointer< ThreadHandle > Thread_Mem( Thread_Number );
if( !Thread_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
ThreadHandle* Threads = Thread_Mem.Get_Mem();
uint8_t i( 0 );
Smart_Array_Pointer< Little_Thread_Helper > Thread_Helper_Mem( Thread_Number );
if( !Thread_Helper_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
Little_Thread_Helper* Thread_Helpers = Thread_Helper_Mem.Get_Mem();
for( ; i < Thread_Number; ++i )
Thread_Helpers[i].m_Blocker = &Blocker; //That isn't that nice...but needed
for( i=0; i < Thread_Number; ++i )
ThreadCreate( Threads[i], Memory_DA_Test_Thread<Buf_Size,Loops>, &Thread_Helpers[i] );
//Now it's starts!
const double Start_Time = Stopwatch::Get_Time();
Blocker = false;
ThreadWaitFor( Threads, Thread_Number );
if( m_Error )
return 0;
for( i=0; i < Thread_Number; ++i )
if( Comp_Time < Thread_Helpers[i].m_Needed_Time - Start_Time )
Comp_Time = Thread_Helpers[i].m_Needed_Time - Start_Time;
return ((Loops*Thread_Number) / Comp_Time)/1000;
}
else
{
watch.Start();
if( !Intern_Mem_DA_Test< Buf_Size, Loops > () )
{
m_Error = Benchmark::Mem_Alloc;
return 0;
}
watch.Stop( &Comp_Time );
return (Loops / Comp_Time)/1000;
}
}
Benchmark::Test_Re_Type Benchmark::Calculation_Test()
{
Stopwatch watch;
double Comp_Time(0);
if( !::Force_No_Threads )
{
bool Blocker( true );
Smart_Array_Pointer< ThreadHandle > Threads_Mem( Thread_Number );
if( !Threads_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
ThreadHandle* Threads = Threads_Mem.Get_Mem();
uint8_t i( 0 );
//Preallocates us the memory needed..
Smart_Array_Pointer< Thread_Helper< Type, Loops > > Thread_Helpers_Mem( Thread_Number );
if( !Thread_Helpers_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
Thread_Helper< Type, Loops >* Thread_Helpers = Thread_Helpers_Mem.Get_Mem();
for( ; i < Thread_Number; ++i )
Thread_Helpers[i].m_Blocker = &Blocker; //That isn't that nice...but needed
for( i=0; i < Thread_Number; ++i )
ThreadCreate( Threads[i], Calculation_Test_Thread< Type, Loops >, &Thread_Helpers[i] );
//Now it's starts!
//Maybe we should wait a moment, so every
//watch.Start();
const double Start_Time = Stopwatch::Get_Time();
Blocker = false;
ThreadWaitFor( Threads, Thread_Number );
for( i=0; i < Thread_Number; ++i )
if( Comp_Time < Thread_Helpers[i].m_Needed_Time - Start_Time )
Comp_Time = Thread_Helpers[i].m_Needed_Time - Start_Time;
double Score = static_cast< double > ( ((Loops-1)*(Loops-1)*Thread_Number) / Comp_Time);
return Score / 1000; //A little score modi :)
}
else
{
Smart_Array_Pointer< Type > Buffer_Mem( Loops );
if( !Buffer_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
Type* Buffer = Buffer_Mem.Get_Mem(); //Does we need some extra to prevent for buffer over/under flows?
watch.Start();
Intern_Calculation_Test< Type, Loops-1 > ( Buffer );
watch.Stop( &Comp_Time );
double Score = static_cast< double > ( ((Loops-1)*(Loops-1)) / Comp_Time);
return Score / 1000; //Little score modi!
}
}
Benchmark::Test_Re_Type Benchmark::Memory_RW_Test()
{
Stopwatch watch;
double Comp_Time( 0 );
static const uint32_t Half_Buf_Size( Buf_Size>>1 );
if( !::Force_No_Threads )
{
bool Blocker( true );
Smart_Array_Pointer< ThreadHandle > Thread_Mem( Thread_Number );
if( !Thread_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
ThreadHandle* Threads = Thread_Mem;
uint8_t i( 0 );
Smart_Array_Pointer< Thread_Helper< uint8_t, Buf_Size > > Thread_Helper_Mem( Thread_Number );
if( !Thread_Helper_Mem || !Thread_Helper_Mem.Get_Mem()->m_Mem.Is_Allocated() ) //uah...evil line :x
{
m_Error = Mem_Alloc;
return 0;
}
Thread_Helper< uint8_t, Buf_Size >* Thread_Helpers = Thread_Helper_Mem.Get_Mem();
for( ; i < Thread_Number; ++i )
Thread_Helpers[i].m_Blocker = &Blocker; //That isn't that nice...but needed
for( i=0; i < Thread_Number; ++i )
ThreadCreate( Threads[i], Memory_RW_Test_Thread< Half_Buf_Size, Loops >, &Thread_Helpers[i] );
//Now it's starts!
//Maybe we should wait a moment, so every
//watch.Start();
const double Start_Time = Stopwatch::Get_Time();
Blocker = false;
ThreadWaitFor( Threads, Thread_Number );
//watch.Stop( &Comp_Time );
for( i=0; i < Thread_Number; ++i )
if( Comp_Time < Thread_Helpers[i].m_Needed_Time - Start_Time )
Comp_Time = Thread_Helpers[i].m_Needed_Time - Start_Time;
return ((Loops*Thread_Number) / Comp_Time)/1000;
}
else
{
Smart_Array_Pointer< uint8_t > a( Half_Buf_Size );
Smart_Array_Pointer< uint8_t > b( Half_Buf_Size );
if( !a || !b )
{
m_Error = Mem_Alloc;
return 0;
}
watch.Start();
Intern_Memory_RW_Test< Half_Buf_Size, Loops > ( a.Get_Mem(), b.Get_Mem() );
watch.Stop( &Comp_Time );
return (Loops / Comp_Time)/1000;
}
}
//------------------------------------------------------------------------------
int main(int argc, char ** argv)
{
bool fullscreen = false;
std::string str;
for (int i = 1; i < argc; ++i) {
if (!strcmp(argv[i], "-d"))
g_level = atoi(argv[++i]);
else if (!strcmp(argv[i], "-c"))
g_repeatCount = atoi(argv[++i]);
else if (!strcmp(argv[i], "-f"))
fullscreen = true;
else {
std::ifstream ifs(argv[1]);
if (ifs) {
std::stringstream ss;
ss << ifs.rdbuf();
ifs.close();
str = ss.str();
g_defaultShapes.push_back(SimpleShape(str.c_str(), argv[1], kCatmark));
}
}
}
initializeShapes();
OsdSetErrorCallback(callbackError);
if (not glfwInit()) {
printf("Failed to initialize GLFW\n");
return 1;
}
static const char windowTitle[] = "OpenSubdiv glBatchViewer";
#define CORE_PROFILE
#ifdef CORE_PROFILE
setGLCoreProfile();
#endif
#if GLFW_VERSION_MAJOR>=3
if (fullscreen) {
g_primary = glfwGetPrimaryMonitor();
// apparently glfwGetPrimaryMonitor fails under linux : if no primary,
// settle for the first one in the list
if (not g_primary) {
int count=0;
GLFWmonitor ** monitors = glfwGetMonitors(&count);
if (count)
g_primary = monitors[0];
}
if (g_primary) {
GLFWvidmode const * vidmode = glfwGetVideoMode(g_primary);
g_width = vidmode->width;
g_height = vidmode->height;
}
}
if (not (g_window=glfwCreateWindow(g_width, g_height, windowTitle,
fullscreen and g_primary ? g_primary : NULL, NULL))) {
printf("Failed to open window.\n");
glfwTerminate();
return 1;
}
glfwMakeContextCurrent(g_window);
glfwSetKeyCallback(g_window, keyboard);
glfwSetCursorPosCallback(g_window, motion);
glfwSetMouseButtonCallback(g_window, mouse);
glfwSetWindowSizeCallback(g_window, reshape);
#else
if (glfwOpenWindow(g_width, g_height, 8, 8, 8, 8, 24, 8,
fullscreen ? GLFW_FULLSCREEN : GLFW_WINDOW) == GL_FALSE) {
printf("Failed to open window.\n");
glfwTerminate();
return 1;
}
glfwSetWindowTitle(windowTitle);
glfwSetKeyCallback(keyboard);
glfwSetMousePosCallback(motion);
glfwSetMouseButtonCallback(mouse);
glfwSetWindowSizeCallback(reshape);
#endif
#if defined(OSD_USES_GLEW)
#ifdef CORE_PROFILE
// this is the only way to initialize glew correctly under core profile context.
glewExperimental = true;
#endif
if (GLenum r = glewInit() != GLEW_OK) {
printf("Failed to initialize glew. Error = %s\n", glewGetErrorString(r));
exit(1);
}
#ifdef CORE_PROFILE
// clear GL errors which was generated during glewInit()
glGetError();
#endif
#endif
// activate feature adaptive tessellation if OSD supports it
g_adaptive = OpenSubdiv::OsdGLDrawContext::SupportsAdaptiveTessellation();
initGL();
glfwSwapInterval(0);
initHUD();
callbackModel(g_currentShape);
g_fpsTimer.Start();
while (g_running) {
idle();
display();
#if GLFW_VERSION_MAJOR>=3
glfwPollEvents();
glfwSwapBuffers(g_window);
#else
glfwSwapBuffers();
#endif
}
uninitGL();
glfwTerminate();
}
Benchmark::Test_Re_Type Benchmark::MultiT_MLuL_Test()
{
Stopwatch watch;
double Comp_Time( 0 );
if( !::Force_No_Threads )
{
bool Blocker( true );
uint8_t i( 0 );
Smart_Array_Pointer< ThreadHandle > Thread_Mem( Thread_Number );
ThreadHandle*const Threads = Thread_Mem.Get_Mem();
#if IsLin
Smart_Array_Pointer< Little_Mutex_Helper > Mutex_Mem( Thread_Number );
if( !Thread_Mem.Is_Allocated() || !Mutex_Mem.Is_Allocated() )
{
m_Error = Mem_Alloc;
return 0;
}
Little_Mutex_Helper*const Mutex = Mutex_Mem.Get_Mem();
#elif IsWin
Smart_Array_Pointer< CRITICAL_SECTION > CS_Mem( Thread_Number );
CRITICAL_SECTION*const CS = CS_Mem.Get_Mem();
for( ; i < Thread_Number; ++i )
InitializeCriticalSection( &CS[i] );
#endif
//Preallocates us the memory needed...
Smart_Array_Pointer< Special_Thread_Helper > Thread_Helpers_Mem( Thread_Number );
if( !Thread_Helpers_Mem )
{
m_Error = Mem_Alloc;
return 0;
}
Special_Thread_Helper* Thread_Helpers = Thread_Helpers_Mem.Get_Mem();
for( i =0; i < Thread_Number; ++i )
{
Thread_Helpers[i].m_Blocker = &Blocker; //That isn't that nice...but needed
Thread_Helpers[i].m_Mutex =
#if IsLin
&Mutex[i].m_Mutex
#elif IsWin
&CS[i]
#endif
;
}
for( i=0; i < Thread_Number; ++i )
ThreadCreate( Threads[i], MultiT_MLuL_Test_Thread< Loops >, &Thread_Helpers[i] );
//Now it's starts!
//Maybe we should wait a moment, so every
double Start_Time = Stopwatch::Get_Time();
Blocker = false;
ThreadWaitFor( Threads, Thread_Number );
for( i=0; i < Thread_Number; ++i )
if( Comp_Time < Thread_Helpers[i].m_Needed_Time - Start_Time )
Comp_Time = Thread_Helpers[i].m_Needed_Time - Start_Time;
return ((Loops*Thread_Number) / Comp_Time)/1000;
}
else
{
#if IsLin
Little_Mutex_Helper Mutex;
watch.Start();
Intern_MultiT_MLuL_Test< Loops > ( &Mutex.m_Mutex );
#elif IsWin
CRITICAL_SECTION CS; //Is critical section relly the right choice?
InitializeCriticalSection( &CS );
watch.Start();
Intern_MultiT_MLuL_Test< Loops > ( &CS );
#endif
watch.Stop( &Comp_Time );
return (Loops / Comp_Time)/1000;
}
}
//todo: Use a typedef for the saving type
void Benchmark::Run()
{
Stopwatch watch;
uint32_t tmp_Time;
if( ::Just_Mem_Test )
tmp_Time = Time_Per_Test * 5;
else if( ::Just_Cpu_Test )
tmp_Time = Time_Per_Test * 7;
else if( ::Just_MT_Test )
tmp_Time = Time_Per_Test * 1;
else
tmp_Time = Time_Per_Test * 13;
if( No_Cpu_Test )
tmp_Time -= Time_Per_Test * 7;
if( No_Mem_Test )
tmp_Time -= Time_Per_Test * 5;
if( No_MT_Test )
tmp_Time -= Time_Per_Test * 1;
const uint32_t Time( tmp_Time );
uint32_t Loops(0);
double Re8int(0),
Re16int(0),
Re32int(0),
Re64int(0),
Re32Fp(0),
Re64Fp(0),
Re96Fp(0),
Re128Fp(0),
ReMemRW2(0),
ReMemRW4(0),
ReMemDA2(0),
ReMemDA4(0),
ReMemDA8(0),
ReMTMLuL(0);
double Cur_Time(0);
Status_Done(0);
watch.Start();
for( ; Cur_Time < Time; ++Loops, watch.Look( Cur_Time ), Status_Done( static_cast< uint32_t >( (100.0/Time)*Cur_Time )) )
{
if( !Just_Mem_Test && !Just_MT_Test && !No_Cpu_Test)
{
Re8int += static_cast< double > (Intern_Integer8_Test() );
if( m_Error )
return;
Re16int += static_cast< double > (Intern_Integer16_Test() );
if( m_Error )
return;
Re32int += static_cast< double > (Intern_Integer32_Test() );
if( m_Error )
return;
Re64int += static_cast< double > (Intern_Integer64_Test() );
Re32Fp += static_cast< double > (Intern_Fp32_Test() );
if( m_Error )
return;
Re64Fp += static_cast< double > (Intern_Fp64_Test() );
if( m_Error )
return;
if( sizeof( long double ) == 12 )
Re96Fp += static_cast< double > (Intern_Fp96_Test() );
else if( sizeof( long double ) == 16 )
Re128Fp += static_cast< double > (Intern_Fp128_Test() );
if( m_Error )
return;
}
if( !Just_Cpu_Test && !Just_MT_Test && !No_Mem_Test)
{
ReMemRW2 += static_cast< double > (Intern_Mem_RW2_Test() );
if( m_Error )
return;
ReMemRW4 += static_cast< double > (Intern_Mem_RW4_Test() );
if( m_Error )
return;
ReMemDA2 += static_cast< double > (Intern_Mem_DA2_Test() );
if( m_Error )
return;
ReMemDA8 += static_cast< double > (Intern_Mem_DA8_Test() );
if( m_Error )
return;
ReMemDA4 += static_cast< double > (Intern_Mem_DA4_Test() );
if( m_Error )
return;
}
if( !Just_Cpu_Test && !Just_Mem_Test && !No_MT_Test )
{
ReMTMLuL += static_cast< double > (Intern_MT_MLuL_Test() );
if( m_Error )
return;
}
}
if( !Just_Mem_Test && !Just_MT_Test && !No_Cpu_Test)
{
//Todo:Change the cast to use the typedef!
m_Integer8_Score = static_cast< Benchmark::Score_Type > ( (Re8int / static_cast<double> (Loops) ) );
m_Integer16_Score = static_cast< Benchmark::Score_Type > ( (Re16int / static_cast<double> (Loops) ) );
m_Integer32_Score = static_cast< Benchmark::Score_Type > ( (Re32int / Loops ) );
m_Integer64_Score = static_cast< Benchmark::Score_Type > ( (Re64int / static_cast<double> (Loops) ) );
m_Fp32_Score = static_cast< Benchmark::Score_Type > ( (Re32Fp / static_cast<double> (Loops) ) );
m_Fp64_Score = static_cast< Benchmark::Score_Type > ( (Re64Fp / static_cast<double> (Loops) ) );
m_Fp96_Score = static_cast< Benchmark::Score_Type > ( (Re96Fp / static_cast<double> (Loops) ) );
m_Fp128_Score = static_cast< Benchmark::Score_Type > ( (Re128Fp / static_cast<double>(Loops) ) );
}
if( !Just_Cpu_Test && !Just_MT_Test && !No_Mem_Test)
{
m_Mem_RW2_Score = static_cast< Benchmark::Score_Type > ( ReMemRW2 / Loops );
m_Mem_RW4_Score = static_cast< Benchmark::Score_Type > ( ReMemRW4 / Loops );
m_Mem_DA2_Score = static_cast< Benchmark::Score_Type > ( ReMemDA2 / Loops );
m_Mem_DA4_Score = static_cast< Benchmark::Score_Type > ( ReMemDA4 / Loops );
m_Mem_DA8_Score = static_cast< Benchmark::Score_Type > ( ReMemDA8 / Loops );
}
if( !Just_Cpu_Test && ! Just_Mem_Test && !No_MT_Test)
{
m_MT_MLuL_Score = static_cast< Benchmark::Score_Type > ( ReMTMLuL / Loops );
}
}
Awaitable<void> ReplicatorPerfTest::Run(
__in wstring const & testFolder,
__in int concurrentTransactions,
__in int totalTransactions,
__in Data::Log::LogManager & logManager)
{
#ifndef PERF_TEST
UNREFERENCED_PARAMETER(testFolder);
UNREFERENCED_PARAMETER(concurrentTransactions);
UNREFERENCED_PARAMETER(totalTransactions);
UNREFERENCED_PARAMETER(logManager);
#else
Replica::SPtr replica = Replica::Create(
pId_,
rId_,
testFolder,
logManager,
underlyingSystem_->PagedAllocator());
co_await replica->OpenAsync();
FABRIC_EPOCH epoch1; epoch1.DataLossNumber = 1; epoch1.ConfigurationNumber = 1; epoch1.Reserved = nullptr;
co_await replica->ChangeRoleAsync(epoch1, FABRIC_REPLICA_ROLE_PRIMARY);
replica->SetReadStatus(FABRIC_SERVICE_PARTITION_ACCESS_STATUS_GRANTED);
replica->SetWriteStatus(FABRIC_SERVICE_PARTITION_ACCESS_STATUS_GRANTED);
KUri::CSPtr stateProviderName = GetStateProviderName(0);
{
Transaction::SPtr txn;
replica->TxnReplicator->CreateTransaction(txn);
KFinally([&] {txn->Dispose(); });
NTSTATUS status = co_await replica->TxnReplicator->AddAsync(*txn, *stateProviderName, L"ReplicatorPerfTest");
VERIFY_IS_TRUE(NT_SUCCESS(status));
co_await txn->CommitAsync();
}
{
IStateProvider2::SPtr stateProvider2;
NTSTATUS status = replica->TxnReplicator->Get(*stateProviderName, stateProvider2);
VERIFY_IS_TRUE(NT_SUCCESS(status));
VERIFY_IS_NOT_NULL(stateProvider2);
VERIFY_ARE_EQUAL(*stateProviderName, stateProvider2->GetName());
IStore<int, int>::SPtr store = dynamic_cast<IStore<int, int>*>(stateProvider2.RawPtr());
Stopwatch s;
s.Start();
KArray<Awaitable<void>> tasks(underlyingSystem_->PagedAllocator(), concurrentTransactions, 0);
for (int i = 0; i < concurrentTransactions; i++)
{
status = tasks.Append(DoWorkOnKey(store, replica, totalTransactions / concurrentTransactions, i));
KInvariant(NT_SUCCESS(status));
}
co_await TaskUtilities<Awaitable<void>>::WhenAll(tasks);
s.Stop();
int64 txPerSec = ((totalTransactions * 1000) / s.ElapsedMilliseconds);
Trace.WriteInfo(
TraceComponent,
"{0}: Tx/Sec is {1}",
prId_->TraceId,
txPerSec);
}
replica->SetReadStatus(FABRIC_SERVICE_PARTITION_ACCESS_STATUS_NOT_PRIMARY);
replica->SetWriteStatus(FABRIC_SERVICE_PARTITION_ACCESS_STATUS_NOT_PRIMARY);
co_await replica->CloseAsync();
#endif
co_return;
}
//------------------------------------------------------------------------------
static void
display() {
// set effect
g_effect.displayStyle = g_displayStyle;
g_effect.screenSpaceTess = (g_screenSpaceTess != 0);
g_effect.displayPatchColor = (g_displayPatchColor != 0);
// prepare view matrix
float aspect = g_width/(float)g_height;
float modelview[16], projection[16];
identity(modelview);
translate(modelview, -g_pan[0], -g_pan[1], -g_dolly);
rotate(modelview, g_rotate[1], 1, 0, 0);
rotate(modelview, g_rotate[0], 0, 1, 0);
rotate(modelview, -90, 1, 0, 0);
translate(modelview, -g_center[0], -g_center[1], -g_center[2]);
perspective(projection, 45.0f, aspect, 0.01f, 500.0f);
g_effect.SetMatrix(modelview, projection);
g_effect.SetTessLevel((float)(1 << g_tessLevel));
g_effect.SetLighting();
// -----------------------------------------------------------------------
// prepare draw items
Stopwatch s;
s.Start();
OpenSubdiv::OsdUtilDrawItem<MyDrawDelegate::EffectHandle, MyDrawContext>::Collection items;
OpenSubdiv::OsdUtilDrawItem<MyDrawDelegate::EffectHandle, MyDrawContext>::Collection cachedDrawItems;
int numModels = g_modelCount*g_modelCount;
items.reserve(numModels);
for (int i = 0; i < numModels; ++i) {
// Here, client can pack arbitrary mesh and effect into drawItems.
items.push_back(OpenSubdiv::OsdUtilDrawItem<MyDrawDelegate::EffectHandle, MyDrawContext>(
g_batch, &g_effect, g_batch->GetPatchArrays(i)));
}
if (g_batching) {
// create cached draw items
OpenSubdiv::OsdUtil::OptimizeDrawItem(items, cachedDrawItems, &g_drawDelegate);
}
s.Stop();
float prepCpuTime = float(s.GetElapsed() * 1000.0f);
// -----------------------------------------------------------------------
// draw items
s.Start();
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glViewport(0, 0, g_width, g_height);
// primitive counting
glBeginQuery(GL_PRIMITIVES_GENERATED, g_queries[0]);
#if defined(GL_VERSION_3_3)
glBeginQuery(GL_TIME_ELAPSED, g_queries[1]);
#endif
g_drawDelegate.ResetNumDrawCalls();
if (g_displayStyle == kWire) glDisable(GL_CULL_FACE);
if (g_batching) {
OpenSubdiv::OsdUtil::DrawCollection(cachedDrawItems, &g_drawDelegate);
} else {
OpenSubdiv::OsdUtil::DrawCollection(items, &g_drawDelegate);
}
if (g_displayStyle == kWire) glEnable(GL_CULL_FACE);
glEndQuery(GL_PRIMITIVES_GENERATED);
#if defined(GL_VERSION_3_3)
glEndQuery(GL_TIME_ELAPSED);
#endif
s.Stop();
float drawCpuTime = float(s.GetElapsed() * 1000.0f);
// -----------------------------------------------------------------------
GLuint numPrimsGenerated = 0;
GLuint timeElapsed = 0;
glGetQueryObjectuiv(g_queries[0], GL_QUERY_RESULT, &numPrimsGenerated);
#if defined(GL_VERSION_3_3)
glGetQueryObjectuiv(g_queries[1], GL_QUERY_RESULT, &timeElapsed);
#endif
float drawGpuTime = timeElapsed / 1000.0f / 1000.0f;
if (g_hud.IsVisible()) {
g_fpsTimer.Stop();
g_totalTime += g_fpsTimer.GetElapsed();
double fps = 1.0/g_fpsTimer.GetElapsed();
g_fpsTimer.Start();
g_hud.DrawString(10, -200, "Draw Calls : %d", g_drawDelegate.GetNumDrawCalls());
g_hud.DrawString(10, -180, "Tess level : %d", g_tessLevel);
g_hud.DrawString(10, -160, "Primitives : %d", numPrimsGenerated);
g_hud.DrawString(10, -120, "GPU Kernel : %.3f ms", g_gpuTime);
g_hud.DrawString(10, -100, "CPU Kernel : %.3f ms", g_cpuTime);
g_hud.DrawString(10, -80, "GPU Draw : %.3f ms", drawGpuTime);
g_hud.DrawString(10, -60, "CPU Draw : %.3f ms", drawCpuTime);
g_hud.DrawString(10, -40, "CPU Prep : %.3f ms", prepCpuTime);
g_hud.DrawString(10, -20, "FPS : %3.1f", fps);
g_hud.Flush();
}
checkGLErrors("display leave");
glFinish();
}