Test hemm no yield async rw mutex

ci/.gitlab-ci.yml

@@ -1,19 +1,19 @@
 include:
-  - local: 'ci/cpu/asan_ubsan_lsan.yml'
+  # - local: 'ci/cpu/asan_ubsan_lsan.yml'
   - local: 'ci/cpu/clang15_release_cxx20.yml'
   - local: 'ci/cpu/clang15_release_stdexec.yml'
   - local: 'ci/cpu/clang15_release.yml'
   - local: 'ci/cpu/clang16_release.yml'
   - local: 'ci/cpu/clang18_release.yml'
   - local: 'ci/cpu/gcc11_release_stdexec.yml'
-  - local: 'ci/cpu/gcc11_debug_stdexec.yml'
+  # - local: 'ci/cpu/gcc11_debug_stdexec.yml'
   - local: 'ci/cpu/gcc12_release_cxx20.yml'
-  - local: 'ci/cpu/gcc13_codecov.yml'
+  # - local: 'ci/cpu/gcc13_codecov.yml'
   - local: 'ci/cpu/gcc13_release.yml'
-  - local: 'ci/cuda/gcc11_release.yml'
-  - local: 'ci/cuda/gcc11_release_scalapack.yml'
-  - local: 'ci/cuda/gcc11_codecov.yml'
-  - local: 'ci/cuda/gcc11_debug_scalapack.yml'
-  - local: 'ci/cuda/gcc13_release_stdexec.yml'
-  - local: 'ci/rocm/clang14_release.yml'
-  - local: 'ci/rocm/clang14_release_stdexec.yml'
+  # - local: 'ci/cuda/gcc11_release.yml'
+  # - local: 'ci/cuda/gcc11_release_scalapack.yml'
+  # - local: 'ci/cuda/gcc11_codecov.yml'
+  # - local: 'ci/cuda/gcc11_debug_scalapack.yml'
+  # - local: 'ci/cuda/gcc13_release_stdexec.yml'
+  # - local: 'ci/rocm/clang14_release.yml'
+  # - local: 'ci/rocm/clang14_release_stdexec.yml'

ci/ctest_to_gitlab.sh

@@ -61,7 +61,7 @@ JOB_TEMPLATE="
     USE_MPI: 'YES'
     DISABLE_AFTER_SCRIPT: 'YES'
     DLAF_HDF5_TEST_OUTPUT_PATH: \$CI_PROJECT_DIR
-  script: mpi-ctest -L {{LABEL}}
+  script: mpi-ctest --timeout 300 --repeat-until-fail 50 -R test_multiplication_hermitian -L {{LABEL}}
   artifacts:
     paths:
       - codecov-reports/"
@@ -100,7 +100,7 @@ JOB_TEMPLATE="
     USE_MPI: 'YES'
     DISABLE_AFTER_SCRIPT: 'YES'
     DLAF_HDF5_TEST_OUTPUT_PATH: \$CI_PROJECT_DIR
-  script: mpi-ctest -L {{LABEL}}"
+  script: mpi-ctest --timeout 60 --repeat-until-fail 50 -R test_multiplication_hermitian -L {{LABEL}}"
 fi
 
 JOBS=""

ci/docker/common.yaml

@@ -46,3 +46,6 @@ packages:
     # Force git as non-buildable to allow deprecated versions in environments
     # https://github.com/spack/spack/pull/30040
     buildable: false
+  pika:
+    require:
+      - '@git.no-yield-async-rw-mutex=0.29.0'

test/unit/multiplication/test_multiplication_hermitian.cpp

@@ -99,8 +99,16 @@ void testHermitianMultiplication(const blas::Side side, const blas::Uplo uplo, c
   // SCOPED_TRACE cannot yield.
   mat_ch.waitLocalTiles();
   SCOPED_TRACE(::testing::Message() << "m " << m << "n " << n << ", mb " << mb << ", nb " << nb);
+  auto t_before = pika::get_worker_thread_num();
   CHECK_MATRIX_NEAR(res_c, mat_ch, 10 * (m + 1) * TypeUtilities<T>::error,
                     10 * (m + 1) * TypeUtilities<T>::error);
+  auto t_after = pika::get_worker_thread_num();
+  if (t_before != t_after) {
+    std::cerr << "worker thread index changed from " << t_before << " to " << t_after << std::endl;
+  }
+  else {
+    std::cerr << "worker thread index is still " << t_before << std::endl;
+  }
 }
 
 template <class T, Backend B, Device D>
@@ -154,70 +162,79 @@ TYPED_TEST(HermitianMultiplicationTestMC, CorrectnessLocal) {
       if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
         continue;
       for (const auto& [m, n, mb, nb] : sizes) {
+        std::cerr << "side: " << side;
+        std::cerr << ", uplo: " << uplo;
+        std::cerr << ", m: " << m;
+        std::cerr << ", n: " << n;
+        std::cerr << ", mb: " << mb;
+        std::cerr << ", nb: " << nb << std::endl;
         TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
         TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
 
         testHermitianMultiplication<TypeParam, Backend::MC, Device::CPU>(side, uplo, m, n, mb, nb, alpha,
                                                                          beta);
+        pika::wait();
+        std::cerr << "done\n";
       }
     }
   }
 }
 
-TYPED_TEST(HermitianMultiplicationTestMC, CorrectnessDistributed) {
-  for (auto& comm_grid : this->commGrids()) {
-    for (const auto side : blas_sides) {
-      for (const auto uplo : blas_uplos) {
-        if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
-          continue;
-
-        for (const auto& [m, n, mb, nb] : sizes) {
-          TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
-          TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
-
-          testHermitianMultiplication<TypeParam, Backend::MC, Device::CPU>(comm_grid, side, uplo, m, n,
-                                                                           mb, nb, alpha, beta);
-          pika::wait();
-        }
-      }
-    }
-  }
-}
-
-#ifdef DLAF_WITH_GPU
-TYPED_TEST(HermitianMultiplicationTestGPU, CorrectnessLocal) {
-  for (const auto side : blas_sides) {
-    for (const auto uplo : blas_uplos) {
-      if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
-        continue;
-      for (const auto& [m, n, mb, nb] : sizes) {
-        TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
-        TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
-
-        testHermitianMultiplication<TypeParam, Backend::GPU, Device::GPU>(side, uplo, m, n, mb, nb,
-                                                                          alpha, beta);
-      }
-    }
-  }
-}
-
-TYPED_TEST(HermitianMultiplicationTestGPU, CorrectnessDistributed) {
-  for (auto& comm_grid : this->commGrids()) {
-    for (const auto side : blas_sides) {
-      for (const auto uplo : blas_uplos) {
-        if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
-          continue;
-
-        for (const auto& [m, n, mb, nb] : sizes) {
-          TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
-          TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
-
-          testHermitianMultiplication<TypeParam, Backend::GPU, Device::GPU>(comm_grid, side, uplo, m, n,
-                                                                            mb, nb, alpha, beta);
-          pika::wait();
-        }
-      }
-    }
-  }
-}
-#endif
+// TYPED_TEST(HermitianMultiplicationTestMC, CorrectnessDistributed) {
+//   for (auto& comm_grid : this->commGrids()) {
+//     for (const auto side : blas_sides) {
+//       for (const auto uplo : blas_uplos) {
+//         if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
+//           continue;
+
+//         for (const auto& [m, n, mb, nb] : sizes) {
+//           TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
+//           TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
+
+//           testHermitianMultiplication<TypeParam, Backend::MC, Device::CPU>(comm_grid, side, uplo, m, n,
+//                                                                            mb, nb, alpha, beta);
+//           pika::wait();
+//         }
+//       }
+//     }
+//   }
+// }
+
+// #ifdef DLAF_WITH_GPU
+// TYPED_TEST(HermitianMultiplicationTestGPU, CorrectnessLocal) {
+//   for (const auto side : blas_sides) {
+//     for (const auto uplo : blas_uplos) {
+//       if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
+//         continue;
+//       for (const auto& [m, n, mb, nb] : sizes) {
+//         TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
+//         TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
+
+//         testHermitianMultiplication<TypeParam, Backend::GPU, Device::GPU>(side, uplo, m, n, mb, nb,
+//                                                                           alpha, beta);
+//         pika::wait();
+//       }
+//     }
+//   }
+// }
+
+// TYPED_TEST(HermitianMultiplicationTestGPU, CorrectnessDistributed) {
+//   for (auto& comm_grid : this->commGrids()) {
+//     for (const auto side : blas_sides) {
+//       for (const auto uplo : blas_uplos) {
+//         if (side != blas::Side::Left || uplo != blas::Uplo::Lower)
+//           continue;
+
+//         for (const auto& [m, n, mb, nb] : sizes) {
+//           TypeParam alpha = TypeUtilities<TypeParam>::element(-1.2, .7);
+//           TypeParam beta = TypeUtilities<TypeParam>::element(1.12, -.1);
+
+//           testHermitianMultiplication<TypeParam, Backend::GPU, Device::GPU>(comm_grid, side, uplo, m, n,
+//                                                                             mb, nb, alpha, beta);
+//           pika::wait();
+//         }
+//       }
+//     }
+//   }
+// }
+// #endif