void
time_parallel_for ()
{
std::cout << "\nTiming how long it takes to run parallel_for:\n";
std::cout << "threads\ttime (best of " << ntrials << ")\n";
std::cout << "-------\t----------\n";
for (int i = 0; threadcounts[i] <= numthreads; ++i) {
int nt = wedge ? threadcounts[i] : numthreads;
int its = iterations/nt;
// make a lambda function that spawns a bunch of threads, calls a
// trivial function, then waits for them to finish and tears down
// the group.
auto func = [=](){
parallel_for (0, nt, [](int64_t i){ /*empty*/ });
};
double range;
double t = time_trial (func, ntrials, its, &range);
std::cout << Strutil::format ("%2d\t%5.1f launch %8.1f threads/sec\n",
nt, t, (nt*its)/t);
if (! wedge)
break; // don't loop if we're not wedging
}
}
int
main(int argc, const char **argv)
{
cpInitChipmunk();
cp_collision_slop = 0.2f;
#ifdef TIME_TRIAL
sleep(1);
for(int i=0; i<demoCount; i++){
if(i == 'l' - 'a') continue;
time_trial(i, 1000);
}
// time_trial(0, 1000);
exit(0);
#endif
mouseBody = cpBodyNew(INFINITY, INFINITY);
glutStuff(argc, argv);
runDemo(demos[firstDemoIndex]);
glutMainLoop();
return 0;
}
int main (int argc, char *argv[])
{
#if !defined(NDEBUG) || defined(OIIO_CI) || defined(OIIO_CODECOV)
// For the sake of test time, reduce the default iterations for DEBUG,
// CI, and code coverage builds. Explicit use of --iters or --trials
// will override this, since it comes before the getargs() call.
iterations /= 10;
ntrials = 1;
#endif
getargs (argc, argv);
std::cout << "hw threads = " << Sysutil::hardware_concurrency() << "\n";
std::cout << "threads\ttime (best of " << ntrials << ")\n";
std::cout << "-------\t----------\n";
static int threadcounts[] = { 1, 2, 4, 8, 12, 16, 20, 24, 28, 32, 64, 128, 1024, 1<<30 };
for (int i = 0; threadcounts[i] <= numthreads; ++i) {
int nt = wedge ? threadcounts[i] : numthreads;
int its = iterations/nt;
double range;
double t = time_trial (std::bind(test_atomics,nt,its),
ntrials, &range);
std::cout << Strutil::format ("%2d\t%5.1f range %.2f\t(%d iters/thread)\n",
nt, t, range, its);
if (! wedge)
break; // don't loop if we're not wedging
}
return unit_test_failures;
}
void
time_thread_group ()
{
std::cout << "\nTiming how long it takes to start/end thread_group:\n";
std::cout << "threads\ttime (best of " << ntrials << ")\n";
std::cout << "-------\t----------\n";
for (int i = 0; threadcounts[i] <= numthreads; ++i) {
int nt = wedge ? threadcounts[i] : numthreads;
int its = iterations/nt;
// make a lambda function that spawns a bunch of threads, calls a
// trivial function, then waits for them to finish and tears down
// the group.
auto func = [=](){
thread_group g;
for (int i = 0; i < nt; ++i)
g.create_thread (do_nothing, i);
g.join_all ();
};
double range;
double t = time_trial (func, ntrials, its, &range);
std::cout << Strutil::format ("%2d\t%5.1f launch %8.1f threads/sec\n",
nt, t, (nt*its)/t);
if (! wedge)
break; // don't loop if we're not wedging
}
}
void
time_thread_pool ()
{
std::cout << "\nTiming how long it takes to launch from thread_pool:\n";
std::cout << "threads\ttime (best of " << ntrials << ")\n";
std::cout << "-------\t----------\n";
thread_pool *pool (default_thread_pool());
for (int i = 0; threadcounts[i] <= numthreads; ++i) {
int nt = wedge ? threadcounts[i] : numthreads;
pool->resize (nt);
int its = iterations/nt;
// make a lambda function that spawns a bunch of threads, calls a
// trivial function, then waits for them to finish and tears down
// the group.
auto func = [=](){
task_set<void> taskset (pool);
for (int i = 0; i < nt; ++i) {
taskset.push (pool->push (do_nothing));
}
taskset.wait();
};
double range;
double t = time_trial (func, ntrials, its, &range);
std::cout << Strutil::format ("%2d\t%5.1f launch %8.1f threads/sec\n",
nt, t, (nt*its)/t);
if (! wedge)
break; // don't loop if we're not wedging
}
}
int
main(int argc, const char **argv)
{
ChipmunkDemo demo_list[] = {
LogoSmash,
PyramidStack,
Plink,
BouncyHexagons,
Tumble,
PyramidTopple,
Planet,
Springies,
Pump,
TheoJansen,
Query,
OneWay,
Joints,
Tank,
Chains,
Crane,
ContactGraph,
Buoyancy,
Player,
Slice,
Convex,
Unicycle,
};
demos = demo_list;
demoCount = sizeof(demo_list)/sizeof(ChipmunkDemo);
int trial = 0;
for(int i=0; i<argc; i++){
if(strcmp(argv[i], "-bench") == 0){
demos = bench_list;
demoCount = bench_count;
} else if(strcmp(argv[i], "-trial") == 0){
trial = 1;
}
}
if(trial){
// sleep(1);
for(int i=0; i<demoCount; i++) time_trial(i, 1000);
// time_trial('d' - 'a', 10000);
exit(0);
} else {
mouseBody = cpBodyNew(INFINITY, INFINITY);
glutStuff(argc, argv);
runDemo(demoIndex);
glutMainLoop();
}
return 0;
}
int
main (int argc, char **argv)
{
getargs (argc, argv);
if (input_filename.empty()) {
std::cout << "Error: Must supply a filename.\n";
return -1;
}
imagecache = ImageCache::create ();
imagecache->attribute ("forcefloat", 1);
// Allocate a buffer big enough (for floats)
bool ok = imagecache->get_imagespec (input_filename, spec);
ASSERT (ok);
imagecache->invalidate_all (true); // Don't hold anything
buffer.resize (spec.image_pixels()*spec.nchannels*sizeof(float), 0);
{
double t = time_trial (time_read_image, ntrials);
std::cout << "image_read speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
{
double t = time_trial (time_read_imagebuf, ntrials);
std::cout << "ImageBuf read speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
{
double t = time_trial (time_ic_get_pixels, ntrials);
std::cout << "ImageCache get_pixels speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
imagecache->invalidate_all (true); // Don't hold anything
ImageCache::destroy (imagecache);
return unit_test_failures;
}
void benchmark_convert_type ()
{
const size_t size = 10000000;
const S testval(1.0);
std::vector<S> svec (size, testval);
std::vector<D> dvec (size);
std::cout << Strutil::format("Benchmark conversion of %6s -> %6s : ",
TypeDesc(BaseTypeFromC<S>::value),
TypeDesc(BaseTypeFromC<D>::value));
float time = time_trial (bind (do_convert_type<S,D>, OIIO::cref(svec), OIIO::ref(dvec)),
ntrials, iterations) / iterations;
std::cout << Strutil::format ("%7.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
D r = convert_type<S,D>(testval);
OIIO_CHECK_EQUAL (dvec[size-1], r);
}
int main (int argc, char *argv[])
{
#if !defined(NDEBUG) || defined(OIIO_CI) || defined(OIIO_CODECOV)
// For the sake of test time, reduce the default iterations for DEBUG,
// CI, and code coverage builds. Explicit use of --iters or --trials
// will override this, since it comes before the getargs() call.
iterations /= 10;
ntrials = 1;
#endif
getargs (argc, argv);
double t;
std::cout << "All times are seconds per " << iterations << " bytes.\n\n";
t = time_trial (test_bjhash_small, ntrials);
std::cout << "BJ hash of small data as bytes: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (test_bjhash_small_words, ntrials);
std::cout << "BJ hash of small data as words: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (test_bjhash_big, ntrials);
std::cout << "BJ hash of big data: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (test_bjhash_big_words, ntrials);
std::cout << "BJ hash of big data as words: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_xxhash, &data[0], 2*sizeof(data[0])), ntrials);
std::cout << "XX hash of small data: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_xxhash, &data[0], data.size()*sizeof(data[0])), ntrials);
std::cout << "XX hash of big data: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_bjstrhash, shortstring), ntrials);
std::cout << "BJ strhash hash of short string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_bjstrhash, medstring), ntrials);
std::cout << "BJ strhash hash of medium string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_bjstrhash, longstring), ntrials);
std::cout << "BJ strhash hash of long string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_farmhashstr, shortstring), ntrials);
std::cout << "farmhash of short string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_farmhashstr, medstring), ntrials);
std::cout << "farmhash of medium string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_farmhashstr, longstring), ntrials);
std::cout << "farmhash of long string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_farmhashchar, shortstring.c_str()), ntrials);
std::cout << "farmhash of short char*: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_farmhashchar, medstring.c_str()), ntrials);
std::cout << "farmhash of medium char*: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_farmhashchar, longstring.c_str()), ntrials);
std::cout << "farmhash of long char*: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_xxhash, shortstring.c_str(), shortstring.length()), ntrials);
std::cout << "xxhash XH64 of short string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_xxhash, medstring.c_str(), medstring.length()), ntrials);
std::cout << "xxhash XH64 of medium string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
t = time_trial (boost::bind(test_xxhash, longstring.c_str(), longstring.length()), ntrials);
std::cout << "xxhash XH64 of long string: "
<< Strutil::timeintervalformat(t, 2) << "\n";
return unit_test_failures;
}
void
test_compute ()
{
double time;
ROI roi (0, xres, 0, yres, 0, 1, 0, channels);
std::cout << "Test straightforward as 1D array of float: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_arrays, roi), ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
// imgR.write ("ref.exr");
std::cout << "Test array iterated like an image: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_arrays_like_image, roi), ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
std::cout << "Test array iterated like an image, multithreaded: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_arrays_like_image_multithread_wrapper, roi), ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
std::cout << "Test array as 1D, using SIMD: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_arrays_simd4, roi), ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
std::cout << "Test array iterated like an image, using SIMD: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_arrays_like_image_simd, roi), ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
std::cout << "Test array iterated like an image, using SIMD, multithreaded: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_arrays_like_image_simd_multithread_wrapper, roi), ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
std::cout << "Test ImageBufAlgo::mad 1 thread: ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_IBA, roi, 1),
ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
std::cout << "Test ImageBufAlgo::mad multi-thread " << numthreads << ": ";
ImageBufAlgo::zero (imgR);
time = time_trial (OIIO::bind (test_IBA, roi, numthreads),
ntrials, iterations) / iterations;
std::cout << Strutil::format ("%.1f Mvals/sec", (size/1.0e6)/time) << std::endl;
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,0), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,1), 0.25, 0.001);
OIIO_CHECK_EQUAL_THRESH (imgR.getchannel(xres/2,yres/2,0,2), 0.50, 0.001);
}
int main(void)
{
test_suite();
time_trial();
return 0;
}
int
main (int argc, char **argv)
{
getargs (argc, argv);
if (input_filename.empty()) {
std::cout << "Error: Must supply a filename.\n";
return -1;
}
imagecache = ImageCache::create ();
imagecache->attribute ("forcefloat", 1);
// Allocate a buffer big enough (for floats)
bool ok = imagecache->get_imagespec (input_filename, spec, 0, 0, true);
ASSERT (ok);
imagecache->invalidate_all (true); // Don't hold anything
buffer.resize (spec.image_pixels()*spec.nchannels*sizeof(float), 0);
{
double t = time_trial (time_read_image, ntrials);
std::cout << "read_image speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
if (spec.tile_width == 0) {
double t = time_trial (time_read_scanline_at_a_time, ntrials);
std::cout << "read_scanline (1 at a time) speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
if (spec.tile_width == 0) {
double t = time_trial (time_read_64_scanlines_at_a_time, ntrials);
std::cout << "read_scanlines (64 at a time) speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
{
imagecache->invalidate_all (true); // Don't hold anything
double t = time_trial (time_read_imagebuf, ntrials);
std::cout << "ImageBuf read speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
{
imagecache->invalidate_all (true); // Don't hold anything
double t = time_trial (time_ic_get_pixels, ntrials);
std::cout << "ImageCache get_pixels speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
std::cout << "With autotile = 64:\n";
imagecache->attribute ("autotile", 64);
{
imagecache->invalidate_all (true); // Don't hold anything
double t = time_trial (time_read_imagebuf, ntrials);
std::cout << "ImageBuf read speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
{
imagecache->invalidate_all (true); // Don't hold anything
double t = time_trial (time_ic_get_pixels, ntrials);
std::cout << "ImageCache get_pixels speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
std::cout << "With autotile = 64, autoscanline = 1:\n";
imagecache->attribute ("autotile", 64);
imagecache->attribute ("autoscanline", 1);
{
imagecache->invalidate_all (true); // Don't hold anything
double t = time_trial (time_read_imagebuf, ntrials);
std::cout << "ImageBuf read speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
{
imagecache->invalidate_all (true); // Don't hold anything
double t = time_trial (time_ic_get_pixels, ntrials);
std::cout << "ImageCache get_pixels speed: " << Strutil::timeintervalformat(t,2) << "\n";
}
if (verbose)
std::cout << "\n" << imagecache->getstats(2) << "\n";
imagecache->invalidate_all (true); // Don't hold anything
ImageCache::destroy (imagecache);
return unit_test_failures;
}
