stellar-group / tutorials

This repository contains material for HPX tutorials given by members of the STE||AR-Group

STE||AR-Group Tutorials

• Introduction to Parallel Programming with HPX - July 4-5 2019, HLRS Stuttgart
• Task based Programming with HPX - October 5-6 2017, CSCS Lugano
• Introduction to Parallel Programming with HPX - March 9-10 2017, HLRS Stuttgart
• HPX - October 17-18 2016, LBL, Berkley C++ Summit
• Task based Programming with HPX - September 29-30 2016, CSCS Lugano

tutorials's Issues

Cannot run tutorials/examples/00_exercises/solution8.cpp

I built hpx from master and when I try to compile:

https://github.com/STEllAR-GROUP/tutorials/blob/master/examples/00_exercises/solution8.cpp

It fails with:

undefined reference to `hpx::partitioned_vector_partition

so I did the following:

#include <hpx/hpx_main.hpp>
#include <hpx/include/parallel_reduce.hpp>
#include <hpx/include/partitioned_vector.hpp> //note that this changed

HPX_REGISTER_PARTITIONED_VECTOR(double);

int main()
{
  hpx::partitioned_vector<double> v(1000000, 1.0);
  hpx::parallel::reduce(hpx::parallel::execution::par, std::begin(v),
                        std::end(v), 0.0, std::plus<double>());

  return 0;
}

This compiles but when i run the executable, it segfaults.

lbl2016/2.pptx typo on slide 8

In: tutorials/lbl2016/HPX Workshop (Berkeley C++ Summit) - 2.pptx
on slide 8
Missing variable name pt.

packaged_task<F> (func);
auto f = pt.get_future();

Failed to run CUDA examples

Hi, I built HPX with CUDA with the following CMake command:

cmake -GNinja -DHPX_WITH_CUDA=ON -DCMAKE_INSTALL_PREFIX=/path/to/hpx/install -DHPX_WITH_MALLOC=system -DHPX_WITH_CXX2A=ON -DHPX_WITH_MAX_CPU_COUNT=8 -DHPX_WITH_APEX=TRUE -DAPEX_WITH_OTF2=TRUE -DOTF2_ROOT=/path/to/otf2/install -DAPEX_WITH_CUDA=TRUE .. && ninja install

I wasn't able to build any of the CUDA examples. For example when compiling (https://github.com/STEllAR-GROUP/tutorials/blob/master/examples/00_exercises/solution9.cpp), I get the following error:

../solution9.cpp: In function ‘int main()’:
../solution9.cpp:16:42: error: ‘cuda’ in namespace ‘hpx::compute’ does not name a type
     using allocator_type = hpx::compute::cuda::default_allocator;
                                          ^~~~
../solution9.cpp:17:41: error: ‘cuda’ in namespace ‘hpx::compute’ does not name a type
     using executor_type = hpx::compute::cuda::default_executor;
                                         ^~~~
../solution9.cpp:28:19: error: ‘hpx::compute::cuda’ has not been declared
     hpx::compute::cuda::target device;
                   ^~~~
../solution9.cpp:31:5: error: ‘allocator_type’ was not declared in this scope
     allocator_type alloc(device);
     ^~~~~~~~~~~~~~
../solution9.cpp:31:5: note: suggested alternative: ‘fallocate’
     allocator_type alloc(device);
     ^~~~~~~~~~~~~~
     fallocate
../solution9.cpp:32:5: error: ‘executor_type’ was not declared in this scope
     executor_type exec(device);
     ^~~~~~~~~~~~~
../solution9.cpp:41:48: error: type/value mismatch at argument 2 in template parameter list for ‘template<class T, class Allocator> class hpx::compute::vector’
     hpx::compute::vector<double, allocator_type> c(1000000, 2.0, alloc);
                                                ^
../solution9.cpp:41:48: note:   expected a type, got ‘allocator_type’
../solution9.cpp:41:66: error: ‘alloc’ was not declared in this scope
     hpx::compute::vector<double, allocator_type> c(1000000, 2.0, alloc);
                                                                  ^~~~~
../solution9.cpp:41:66: note: suggested alternative: ‘valloc’
     hpx::compute::vector<double, allocator_type> c(1000000, 2.0, alloc);
                                                                  ^~~~~
                                                                  valloc
../solution9.cpp:41:71: error: expression list treated as compound expression in initializer [-fpermissive]
     hpx::compute::vector<double, allocator_type> c(1000000, 2.0, alloc);
                                                                       ^
../solution9.cpp:44:60: error: ‘exec’ was not declared in this scope
     hpx::parallel::reduce(hpx::parallel::execution::par.on(exec), std::begin(v),
                                                            ^~~~
../solution9.cpp:44:60: note: suggested alternative: ‘rexec’
     hpx::parallel::reduce(hpx::parallel::execution::par.on(exec), std::begin(v),
                                                            ^~~~
                                                            rexec
../solution9.cpp:44:78: error: ‘v’ was not declared in this scope
     hpx::parallel::reduce(hpx::parallel::execution::par.on(exec), std::begin(v),
                                                                              ^
ninja: build stopped: subcommand failed.

I modified the code to reflect current HPX changes (like hpx::compute::cuda::default_allocator to hpx::cuda::experimental::default_allocator) but I kept gettiing error: ‘cuda’ is not a member of ‘hpx’.