Build failure on MacOS 14.7.1

[lucbv]

Hi all,
I am trying to build the library on my mac and hit the following build failure:

Building CXX object tests/native/CMakeFiles/gemv_no_ambig.dir/gemv_no_ambig.cpp.o
/Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/gemv_no_ambig.cpp:47:43: error: no member named 'par' in namespace 'std::execution'
   47 |     matrix_vector_product(std::execution::par,
      |                           ~~~~~~~~~~~~~~~~^

My compiler version is:

Apple clang version 16.0.0 (clang-1600.0.26.4)
Target: arm64-apple-darwin23.6.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin

[lucbv]

So to get passed this I simply hard coded #undef LINALG_HAS_EXECUTION in my local copy of stdblas.
Creating a CMake option that allows me to explicitly disable LINALG_HAS_EXECUTION would be helpful.

[lucbv]

A second issue I run into now is the following:

In file included from /Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/iterator.cpp:1:
In file included from /Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/./gtest_fixtures.hpp:21:
In file included from /Users/lberge/Research/kokkos_eco/installs/gtest/include/gtest/gtest.h:55:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/memory:937:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__memory/ranges_uninitialized_algorithms.h:22:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__memory/uninitialized_algorithms.h:13:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/copy.h:12:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/copy_move_common.h:12:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/iterator_operations.h:15:
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__iterator/advance.h:58:7: error: no viable overloaded '+='
   58 |   __i += __n;
      |   ~~~ ^  ~~~
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__iterator/advance.h:71:8: note: in instantiation of function template specialization 'std::__advance<(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>>' requested here
   71 |   std::__advance(__i, __n, typename iterator_traits<_InputIter>::iterator_category());
      |        ^
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__iterator/next.h:35:8: note: in instantiation of function template specialization 'std::advance<(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>, long, long, void>' requested here
   35 |   std::advance(__x, __n);
      |        ^
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/iterator_operations.h:149:17: note: in instantiation of function template specialization 'std::next<(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>, 0>' requested here
  149 |     return std::next(std::forward<_Iter>(__it), __n);
      |                 ^
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/rotate.h:162:31: note: in instantiation of function template specialization 'std::_IterOps<std::_ClassicAlgPolicy>::next<(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>> &>' requested here
  162 |     if (_IterOps<_AlgPolicy>::next(__first) == __middle)
      |                               ^
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/rotate.h:183:24: note: in instantiation of function template specialization 'std::__rotate_impl<std::_ClassicAlgPolicy, (anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>>' requested here
  183 |   auto __result = std::__rotate_impl<_AlgPolicy>(std::move(__first), std::move(__middle), __last_iter, _IterCategory());
      |                        ^
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/rotate.h:191:15: note: in instantiation of function template specialization 'std::__rotate<std::_ClassicAlgPolicy, (anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>, (anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>>' requested here
  191 |   return std::__rotate<_ClassicAlgPolicy>(std::move(__first), std::move(__middle), std::move(__last)).first;
      |               ^
/Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/iterator.cpp:239:10: note: in instantiation of function template specialization 'std::rotate<(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>>' requested here
  239 |     std::rotate(begin(x), begin(x) + 1, end(x));
      |          ^
/Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/iterator.cpp:334:25: note: in instantiation of function template specialization '(anonymous namespace)::testRotateSort<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>' requested here
  334 |     const bool ok_col = testRotateSort(A_col0);
      |                         ^

[mhoemmen]

Hi Luc! Could you try commenting out the definition of LINALG_HAS_EXECUTION on line 53? Lines 46 - 54 of include/experimental/__p1673_bits/macros.hpp would look like this as a result.

#if defined(__NVCOMPILER)
#  define LINALG_HAS_EXECUTION 1
#elif ! defined(__clang__) && ! defined(_MSC_VER) && ! defined(_INTEL_COMPILER) && ! defined(_INTEL_LLMV_COMPILER) && defined(__GNUC__)
#  if defined(LINALG_ENABLE_TBB)
#    define LINALG_HAS_EXECUTION 1
#  endif
#else
// #  define LINALG_HAS_EXECUTION 1
#endif

If that works, then we should be able to figure out a nicer way to express your compiler's needs. Thanks!

[lucbv]

Hi Mark, commenting out the macro definition on line 53 does work as well.
Regarding the issue reported above with the missing operator+= of MdspanRandomAccessIterator , defining the operator in that class make the error disappear, now hitting a new issue:

In file included from /Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/iterator.cpp:2:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/algorithm:1847:
In file included from /Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/nth_element.h:15:
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/sort.h:573:40: error: invalid operands to binary expression ('(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>' and '(anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>')
  573 |   bool __already_partitioned = __first >= __last;
      |                                ~~~~~~~ ^  ~~~~~~
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/sort.h:844:42: note: in instantiation of function template specialization 'std::__bitset_partition<std::_ClassicAlgPolicy, (anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>, std::__less<void, void> &>' requested here
  844 |                                   ? std::__bitset_partition<_AlgPolicy, _RandomAccessIterator, _Compare>(__first, __last, __comp)
      |                                          ^
/Library/Developer/CommandLineTools/SDKs/MacOSX15.1.sdk/usr/include/c++/v1/__algorithm/sort.h:929:8: note: in instantiation of function template specialization 'std::__introsort<std::_ClassicAlgPolicy, std::__less<void, void> &, (anonymous namespace)::MdspanRandomAccessIterator<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>, false>' requested here
  929 |   std::__introsort<_AlgPolicy,
      |        ^
/Users/lberge/Research/kokkos_eco/sources/stdblas/tests/native/iterator.cpp:339:25: note: in instantiation of function template specialization '(anonymous namespace)::testRotateSort<double, Kokkos::extents<unsigned long, 18446744073709551615>, Kokkos::layout_stride, Kokkos::default_accessor<double>>' requested here
  339 |     const bool ok_col = testRotateSort(A_col0);
...

Help with this would also be greatly appreciated : )

[mhoemmen]

I think I know what happened with LINALG_HAS_EXECUTION. Originally there were a few PRs (see e.g., #257 ) that aimed to fix it more comprehensively by adding CMake options (e.g., for enabling TBB). However, we ran out of time and ultimately had to fix use of the macro throughout the code base, without updating the logic for enabling it.

The iterators are completely superfluous; we should actually remove them from the repository.

I don't have a way to test on MacOS right now. Would you consider submitting a PR to fix this issues and verifying the fix locally? Thanks! : - )

[lucbv]

Sorry for the slow response, I started by adding some logic to detect building with apple clang in PR #301

[lucbv]

Now with the addition of PR #303 we can have further confidence in the changes as it adds a MacOS workflow that will test the AppleClang compiler : )