ViewDefaultConstruct( type * pointer , size_t capacity )
: m_ptr( pointer )
{
Kokkos::RangePolicy< ExecSpace > range( 0 , capacity );
parallel_for( range , *this );
ExecSpace::fence();
}
int GhostBlockBrickedVolume::setRegion(
// points to the first voxel to be copied. The voxels at 'source' MUST
// have dimensions 'regionSize', must be organized in 3D-array order, and
// must have the same voxel type as the volume.
const void *source,
// coordinates of the lower, left, front corner of the target region
const vec3i ®ionCoords,
// size of the region that we're writing to, MUST be the same as the
// dimensions of source[][][]
const vec3i ®ionSize)
{
// Create the equivalent ISPC volume container and allocate memory for voxel
// data.
if (ispcEquivalent == nullptr)
createEquivalentISPC();
/*! \todo check if we still need this 'computevoxelrange' - in
theory we need this only if the app is allowed to query these
values, and they're not being set in sharedstructuredvolume,
either, so should we actually set them at all!? */
// Compute the voxel value range for unsigned byte voxels if none was
// previously specified.
Assert2(source,"nullptr source in GhostBlockBrickedVolume::setRegion()");
#ifndef OSPRAY_VOLUME_VOXELRANGE_IN_APP
if (findParam("voxelRange") == NULL) {
// Compute the voxel value range for float voxels if none was
// previously specified.
const size_t numVoxelsInRegion
= (size_t)regionSize.x *
+ (size_t)regionSize.y *
+ (size_t)regionSize.z;
if (voxelType == "uchar")
computeVoxelRange((unsigned char *)source, numVoxelsInRegion);
else if (voxelType == "ushort")
computeVoxelRange((unsigned short *)source, numVoxelsInRegion);
else if (voxelType == "float")
computeVoxelRange((float *)source, numVoxelsInRegion);
else if (voxelType == "double")
computeVoxelRange((double *) source, numVoxelsInRegion);
else {
throw std::runtime_error("invalid voxelType in "
"GhostBlockBrickedVolume::setRegion()");
}
}
#endif
// Copy voxel data into the volume.
const int NTASKS = regionSize.y * regionSize.z;
parallel_for(NTASKS, [&](int taskIndex){
ispc::GBBV_setRegion(ispcEquivalent,
source,
(const ispc::vec3i &)regionCoords,
(const ispc::vec3i &)regionSize,
taskIndex);
});
return true;
}
static void apply( const value_type& alpha ,
const vector_type & x ,
const value_type & beta ,
const vector_type & y )
{
const size_t row_count = x.dimension_0() ;
parallel_for( row_count , Update(alpha,x,beta,y) );
}
__dllexport void ISPCLaunch(void** taskPtr, void* func, void* data, int count)
{
parallel_for(size_t(0), size_t(count),[&] (const range<size_t>& r) {
const size_t threadIndex = tbb::task_arena::current_thread_index();
const size_t threadCount = tbb::task_scheduler_init::default_num_threads();
for (size_t i=r.begin(); i<r.end(); i++) ((TaskFuncType)func)(data,threadIndex,threadCount,i,count);
});
}
void parallel_for(loop_by_eager_binary_splitting<control_by_prediction>& lpalgo,
const Loop_complexity_measure_fct& loop_compl_fct,
Number lo, Number hi, const Body& body) {
auto loop_cutoff_fct = [] (Number lo, Number hi) {
todo();
return false;
};
parallel_for(lpalgo, loop_cutoff_fct, loop_compl_fct, lo, hi, body);
}
void axpby( const ConstScalarType & alpha ,
const ConstVectorType & X ,
const ConstScalarType & beta ,
const VectorType & Y )
{
typedef AXPBY< ConstScalarType , ConstVectorType , VectorType > functor ;
parallel_for( Y.dimension_0() , functor( alpha , X , beta , Y ) );
}
void sample_primary_rays(const Camera &camera,
const BufferView<CameraSample> &samples,
BufferView<Ray> rays,
BufferView<RayDifferential> ray_differentials,
bool use_gpu) {
parallel_for(primary_ray_sampler{
camera, samples.begin(), rays.begin(), ray_differentials.begin()},
samples.size(), use_gpu);
}
Multiply( const matrix_type & A ,
const size_type nrow ,
const size_type , // ncol ,
const vector_type & x ,
const vector_type & y )
: m_A( A ), m_x( x ), m_y( y )
{
parallel_for( nrow , *this );
}
void testThreadedGet()
{
Cache cache( get, hash, 10000, new ObjectPool(10000) );
parallel_for( blocked_range<size_t>( 0, 10000 ), GetFromCache( cache ) );
BOOST_CHECK_EQUAL( size_t(500), cache.cachedComputations() );
}
__dllexport void ISPCLaunch(void** taskPtr, void* func, void* data, int count)
{
parallel_for(size_t(0), size_t(count), [&] (const range<size_t>& r) {
const size_t threadIndex = TaskSchedulerNew::thread()->threadIndex;
const size_t threadCount = TaskSchedulerNew::threadCount();
for (size_t i=r.begin(); i<r.end(); i++)
((TaskFuncType)func)(data,threadIndex,threadCount,i,count);
});
}
extern "C" __dllexport void ISPCLaunch(void** taskPtr, void* func, void* data, int count)
{
parallel_for(0, count,[&] (const range<int>& r) {
const int threadIndex = (int) TaskScheduler::threadIndex();
const int threadCount = (int) TaskScheduler::threadCount();
for (int i=r.begin(); i<r.end(); i++)
((ISPCTaskFunc)func)(data,threadIndex,threadCount,i,count);
});
}
inline static
void apply( const size_t n ,
const double alpha ,
const scalar_vector & w )
{
FILL op ;
op.w = w ;
op.alpha = alpha ;
parallel_for( n , op );
}
void put_all_files(leveldb::DB* db, const std::vector<std::string>& files,
int concurrency) {
auto errors = parallel_for(concurrency, serial_read_files, files, db);
for (bool err : errors)
if (err) {
std::cerr << "Errors occured!" << std::endl;
std::exit(1);
}
}
// dense matrix by dense vector multiplication: d := m * v
// r: # rows in m; c # columns in m
static
void dmdvmult(long r, long c, const float* m, const float* v, float* d) {
auto outer_loop_compl_fct = [c] (long lo, long hi) {
return (hi - lo) * c;
};
parallel_for(outerlp, outer_loop_compl_fct, 0l, r, [&] (long i) {
const float* row_i = &m[i*c];
d[i] = ddotprod(innerlp, r, 0.0f, row_i, v);
});
}
void NativeCurvesISA::commit_helper()
{
if (native_curves.size() != curves.size())
{
native_curves = APIBuffer<unsigned>(scene->device,curves.size(),sizeof(unsigned int),true);
parallel_for(size_t(0), curves.size(), size_t(1024), [&] ( const range<size_t> r) {
for (size_t i=r.begin(); i<r.end(); i++) {
if (curves[i]+3 >= numVertices()) native_curves[i] = 0xFFFFFFF0; // invalid curves stay invalid this way
else native_curves[i] = unsigned(4*i);
}
});
}
if (native_vertices.size() != vertices.size())
native_vertices.resize(vertices.size());
parallel_for(vertices.size(), [&] ( const size_t i ) {
if (native_vertices[i].size() != 4*curves.size())
native_vertices[i] = APIBuffer<Vec3fa>(scene->device,4*curves.size(),sizeof(Vec3fa),true);
parallel_for(size_t(0), curves.size(), size_t(1024), [&] ( const range<size_t> rj ) {
for (size_t j=rj.begin(); j<rj.end(); j++)
{
const unsigned id = curves[j];
if (id+3 >= numVertices()) continue; // ignore invalid curves
const Vec3fa v0 = vertices[i][id+0];
const Vec3fa v1 = vertices[i][id+1];
const Vec3fa v2 = vertices[i][id+2];
const Vec3fa v3 = vertices[i][id+3];
const InputCurve3fa icurve(v0,v1,v2,v3);
OutputCurve3fa ocurve; convert<Vec3fa>(icurve,ocurve);
native_vertices[i].store(4*j+0,ocurve.v0);
native_vertices[i].store(4*j+1,ocurve.v1);
native_vertices[i].store(4*j+2,ocurve.v2);
native_vertices[i].store(4*j+3,ocurve.v3);
}
});
});
native_vertices0 = native_vertices[0];
}
static void apply( const mesh_type & mesh ,
const elem_matrices_type & elem_matrices ,
const elem_vectors_type & elem_vectors ,
const scalar_type elem_coeff_K ,
const scalar_type elem_coeff_Q )
{
ElementComputation comp( mesh , elem_matrices , elem_vectors , elem_coeff_K , elem_coeff_Q );
const size_t elem_count = mesh.elem_node_ids.dimension_0();
parallel_for( elem_count , comp );
}
void
test_thread_pool_recursion ()
{
std::cout << "\nTesting thread pool recursion" << std::endl;
static spin_mutex print_mutex;
thread_pool *pool (default_thread_pool());
pool->resize (2);
parallel_for (0, 10, [&](int id, int64_t i){
// sleep long enough that we can push all the jobs before any get
// done.
Sysutil::usleep (10);
// then run something else that itself will push jobs onto the
// thread pool queue.
parallel_for (0, 10, [&](int id, int64_t i){
Sysutil::usleep (2);
spin_lock lock (print_mutex);
std::cout << " recursive running thread " << id << std::endl;
});
});
}
std::vector<sframe> subplan_executor::run(
const std::vector<std::shared_ptr<planner_node>>& stuff_to_run_in_parallel,
const materialize_options& exec_params) {
std::vector<sframe> ret(stuff_to_run_in_parallel.size());
parallel_for(0, stuff_to_run_in_parallel.size(), [&](const size_t i) {
ret[i] = run(stuff_to_run_in_parallel[i], exec_params);
});
return ret;
}
size_t lualambda_master::make_lambda(const std::string& lambda_str) {
size_t lambda_hash = std::hash<std::string>()(lambda_str);
std::string newstr = lambda_str;
if (boost::starts_with(newstr,"LUA")) {
newstr = newstr.substr(3);
}
parallel_for (0, num_workers(), [&](size_t i) {
clients[i]->doString(newstr);
clients[i]->doString("lambda" + std::to_string(lambda_hash) + " = __lambda__transfer__");
});
return lambda_hash;
}
void FractalRenderer::performRendering(void)
{
printf("data=%p, width=%d, heigth=%d, zoom=%f, resolution=%d, posx=%f, posy=%f\n",
m_data, m_image_x, m_image_y, m_scale, m_resolution, m_normalizedPosition.x, m_normalizedPosition.y);
sf::Clock timer;
parallel_for(tbb::blocked_range2d<unsigned, unsigned>(0, m_image_x, 50, 0, m_image_y, 50),
MandelbrotRenderer(m_data, m_image_x, m_image_y, m_scale, m_resolution, m_normalizedPosition));
m_texture.update(m_data);
m_lastRenderingTime = timer.getElapsedTime();
}
void reduce_floats() {
srand(time(NULL));
const int numStars = 5000;
star* array_of_structs = new star[numStars];
stars struct_of_arrays(numStars);
populate_star_array(array_of_structs, numStars);
populate_stars(&struct_of_arrays);
cout << "map pattern with arrays-of-structures" << endl;
cout << "and structures-of-arrays" << endl;
cout << "----------------------------------------" << endl;
cout << "calculating longitude from right ascension of stars" << endl;
auto get_longitude = [](const double ascension) {
const double degrees = ascension * 15.0;
const double adjusted_degrees = degrees > 360.0 ? degrees - 180.0 : degrees;
return adjusted_degrees;
};
double output[numStars];
const int numIterations = 10;
clock_t start = clock();
for(int i = 0, end = numIterations; i != end; ++i) {
parallel_for(0, numStars, 1, [&struct_of_arrays, &output, get_longitude](int i) {
output[i] = get_longitude(struct_of_arrays.ascension[i]);
});
}
double seconds = (clock() - start) / (double)CLOCKS_PER_SEC;
cout << "structure-of-arrays: " << seconds << " seconds." << endl;
start = clock();
for(int i = 0, end = numIterations; i != end; ++i) {
parallel_for(0, numStars, 1, [&array_of_structs, &output, get_longitude](int i) {
output[i] = get_longitude(array_of_structs[i].ascension);
});
}
seconds = (clock() - start) / (double)CLOCKS_PER_SEC;
cout << "array-of-structures: " << seconds << " seconds." << endl;
}
void assign (size_t n, const Scalar& val) {
/* Resize if necessary (behavour of std:vector) */
if(n>capacity())
DV::resize(size_t (n*_extra_storage));
_size = n;
/* Assign value either on host or on device */
if( DV::modified_host >= DV::modified_device ) {
set_functor_host f(DV::h_view,val);
parallel_for(n,f);
DV::t_host::device_type::fence();
DV::modified_host++;
} else {
set_functor f(DV::d_view,val);
parallel_for(n,f);
DV::t_dev::device_type::fence();
DV::modified_device++;
}
}
int main(int argc, char** argv) {
if(argc != 2) {
printf("Usage: %s <password hash>\n",argv[0]);
return 1;
}
int notfound=1;
SpaceSearcher s( argv[1], ¬found );
parallel_for( tbb::blocked_range<long>( 0, SEARCH_SPACE ), s );
return 0;
}
/* called by the C++ code to render */
extern "C" void device_render (int* pixels,
const unsigned int width,
const unsigned int height,
const float time,
const ISPCCamera& camera)
{
const int numTilesX = (width +TILE_SIZE_X-1)/TILE_SIZE_X;
const int numTilesY = (height+TILE_SIZE_Y-1)/TILE_SIZE_Y;
parallel_for(size_t(0),size_t(numTilesX*numTilesY),[&](const range<size_t>& range) {
for (size_t i=range.begin(); i<range.end(); i++)
renderTileTask((int)i,pixels,width,height,time,camera,numTilesX,numTilesY);
});
}
parray<long> weights(long n, const Weight& weight) {
const long k = 1024;
long m = 1 + ((n - 1) / k);
long tot;
parray<long> rs(n + 1);
#ifdef MANUAL_CONTROL
if (n < WEIGHTS_THRESHOLD) {
tot = weights_seq(weight, 0, n, 0, rs.begin());
rs[n] = tot;
return rs;
}
#endif
par::cstmt(weights_contr<>::weights, [&] { return n; }, [&] {
if (n <= k) {
tot = weights_seq(weight, 0, n, 0, rs.begin());
} else {
parray<long> sums(m);
parallel_for(0l, m, [&] (long i) {
long lo = i * k;
long hi = std::min(lo + k, n);
sums[i] = 0;
for (long j = lo; j < hi; j++) {
sums[i] += weight(j);
}
});
parray<long> scans = rec(sums);
parallel_for(0l, m, [&] (long i) {
long lo = i * k;
long hi = std::min(lo + k, n);
weights_seq(weight, lo, hi, scans[i], rs.begin()+lo);
});
tot = rs[n-1] + weight(n-1);
}
}, [&] {
tot = weights_seq(weight, 0, n, 0, rs.begin());
});
rs[n] = tot;
return rs;
}
inline
static
void unpack( const array_type & arg_output ,
const buffer_type & arg_input ,
const size_type arg_begin ,
const size_type arg_count )
{
UnpackArray op ;
op.output = arg_output ;
op.input = arg_input ;
op.base = arg_begin ;
parallel_for( arg_count , op );
}
void assign_closest(std::vector<IterRange>& parvec,
vuad& totals,
int K,
leveldb::DB* work_db,
const std::vector<GDELTMini>& centroids,
int concurrency,
vuai& cluster_sizes) {
for (auto& i : cluster_sizes) i->store(0);
for (auto& d : totals) d->store(0);
parallel_for(concurrency, assign_closest_serial,
parvec, K, work_db, std::cref(centroids),
std::ref(totals), std::ref(cluster_sizes));
}
ViewRemap( const OutputView & arg_out , const InputView & arg_in )
: output( arg_out ), input( arg_in )
, n0( std::min( (size_t)arg_out.dimension_0() , (size_t)arg_in.dimension_0() ) )
, n1( std::min( (size_t)arg_out.dimension_1() , (size_t)arg_in.dimension_1() ) )
, n2( std::min( (size_t)arg_out.dimension_2() , (size_t)arg_in.dimension_2() ) )
, n3( std::min( (size_t)arg_out.dimension_3() , (size_t)arg_in.dimension_3() ) )
, n4( std::min( (size_t)arg_out.dimension_4() , (size_t)arg_in.dimension_4() ) )
, n5( std::min( (size_t)arg_out.dimension_5() , (size_t)arg_in.dimension_5() ) )
, n6( std::min( (size_t)arg_out.dimension_6() , (size_t)arg_in.dimension_6() ) )
, n7( std::min( (size_t)arg_out.dimension_7() , (size_t)arg_in.dimension_7() ) )
{
parallel_for( n0 , *this );
}
void cv::fastNlMeansDenoisingMulti( InputArrayOfArrays _srcImgs, OutputArray _dst,
int imgToDenoiseIndex, int temporalWindowSize,
float h, int templateWindowSize, int searchWindowSize)
{
std::vector<Mat> srcImgs;
_srcImgs.getMatVector(srcImgs);
fastNlMeansDenoisingMultiCheckPreconditions(
srcImgs, imgToDenoiseIndex,
temporalWindowSize, templateWindowSize, searchWindowSize
);
_dst.create(srcImgs[0].size(), srcImgs[0].type());
Mat dst = _dst.getMat();
switch (srcImgs[0].type()) {
case CV_8U:
parallel_for(cv::BlockedRange(0, srcImgs[0].rows),
FastNlMeansMultiDenoisingInvoker<uchar>(
srcImgs, imgToDenoiseIndex, temporalWindowSize,
dst, templateWindowSize, searchWindowSize, h));
break;
case CV_8UC2:
parallel_for(cv::BlockedRange(0, srcImgs[0].rows),
FastNlMeansMultiDenoisingInvoker<cv::Vec2b>(
srcImgs, imgToDenoiseIndex, temporalWindowSize,
dst, templateWindowSize, searchWindowSize, h));
break;
case CV_8UC3:
parallel_for(cv::BlockedRange(0, srcImgs[0].rows),
FastNlMeansMultiDenoisingInvoker<cv::Vec3b>(
srcImgs, imgToDenoiseIndex, temporalWindowSize,
dst, templateWindowSize, searchWindowSize, h));
break;
default:
CV_Error(CV_StsBadArg,
"Unsupported matrix format! Only uchar, Vec2b, Vec3b are supported");
}
}
void launch_animateSphere(animateSphereFunc func,
int taskSize,
Vertex* vertices,
const float rcpNumTheta,
const float rcpNumPhi,
const Vec3fa& pos,
const float r,
const float f)
{
parallel_for(size_t(0),size_t(taskSize),[&] (const range<size_t>& m) {
for (size_t i=m.begin(); i<m.end(); i++)
func(i,vertices,rcpNumTheta,rcpNumPhi,pos,r,f);
});
}
