Fix setting of global thread number

.github/workflows/linux_valgrind.yml

@@ -85,6 +85,7 @@ jobs:
               tests.unit.modules.string_util \
               tests.{performance,regressions,unit}.modules.synchronization \
               tests.unit.modules.tag_invoke \
+              tests.unit.modules.thread_manager \
               tests.unit.modules.threading \
               tests.{regressions,unit}.modules.threading_base \
               tests.unit.modules.topology \

libs/pika/thread_manager/src/thread_manager.cpp

@@ -11,7 +11,6 @@
 #include <pika/config.hpp>
 #include <pika/assert.hpp>
 #include <pika/execution_base/this_thread.hpp>
-#include <pika/functional/bind.hpp>
 #include <pika/modules/errors.hpp>
 #include <pika/modules/logging.hpp>
 #include <pika/modules/schedulers.hpp>
@@ -60,19 +59,26 @@ namespace pika::threads::detail {
       : rtcfg_(rtcfg)
       , notifier_(notifier)
     {
-        using std::placeholders::_1;
-        using std::placeholders::_3;
-
         // Add callbacks local to thread_manager.
         notifier.add_on_start_thread_callback(
-            util::detail::bind(&thread_manager::init_tss, this, _1));
-        notifier.add_on_stop_thread_callback(util::detail::bind(&thread_manager::deinit_tss, this));
-
-        auto& rp = pika::resource::get_partitioner();
+            [this]([[maybe_unused]] std::size_t local_thread_num, std::size_t global_thread_num,
+                [[maybe_unused]] char const* pool_name,
+                [[maybe_unused]] char const* postfix) { this->init_tss(global_thread_num); });
+        notifier.add_on_stop_thread_callback(
+            [this]([[maybe_unused]] std::size_t local_thread_num,
+                [[maybe_unused]] std::size_t global_thread_num,
+                [[maybe_unused]] char const* pool_name,
+                [[maybe_unused]] char const* postfix) { this->deinit_tss(); });
         notifier.add_on_start_thread_callback(
-            util::detail::bind(&resource::detail::partitioner::assign_pu, std::ref(rp), _3, _1));
+            [](std::size_t local_thread_num, [[maybe_unused]] std::size_t global_thread_num,
+                char const* pool_name, [[maybe_unused]] char const* postfix) {
+                pika::resource::get_partitioner().assign_pu(pool_name, local_thread_num);
+            });
         notifier.add_on_stop_thread_callback(
-            util::detail::bind(&resource::detail::partitioner::unassign_pu, std::ref(rp), _3, _1));
+            [](std::size_t local_thread_num, [[maybe_unused]] std::size_t global_thread_num,
+                char const* pool_name, [[maybe_unused]] char const* postfix) {
+                pika::resource::get_partitioner().unassign_pu(pool_name, local_thread_num);
+            });
     }
 
     void thread_manager::create_pools()

libs/pika/thread_manager/tests/unit/CMakeLists.txt

@@ -1,5 +1,27 @@
-# Copyright (c) 2019 The STE||AR-Group
+# Copyright (c) 2024 ETH Zurich
 #
 # SPDX-License-Identifier: BSL-1.0
 # Distributed under the Boost Software License, Version 1.0. (See accompanying
 # file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
+
+set(tests thread_num)
+
+set(thread_num_PARAMETERS THREADS 4)
+
+foreach(test ${tests})
+  set(sources ${test}.cpp)
+
+  source_group("Source Files" FILES ${sources})
+
+  pika_add_executable(
+    ${test}_test INTERNAL_FLAGS
+    SOURCES ${sources} ${${test}_FLAGS}
+    EXCLUDE_FROM_ALL
+    FOLDER "Tests/Unit/Modules/ThreadManager/"
+  )
+
+  pika_add_unit_test(
+    "modules.thread_manager" ${test} ${${test}_PARAMETERS} VALGRIND
+  )
+
+endforeach()

libs/pika/thread_manager/tests/unit/thread_num.cpp

@@ -0,0 +1,92 @@
+//  Copyright (c) 2017 Mikael Simberg
+//
+//  SPDX-License-Identifier: BSL-1.0
+//  Distributed under the Boost Software License, Version 1.0. (See accompanying
+//  file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
+
+// Simple test verifying basic resource_partitioner functionality.
+
+#include <pika/assert.hpp>
+#include <pika/execution.hpp>
+#include <pika/init.hpp>
+#include <pika/testing.hpp>
+
+#include <cstddef>
+#include <string>
+#include <vector>
+
+namespace ex = pika::execution::experimental;
+namespace tt = pika::this_thread::experimental;
+
+std::size_t const max_threads =
+    (std::min)(std::size_t(4), std::size_t(pika::threads::detail::hardware_concurrency()));
+
+void test_scheduler(int argc, char* argv[], pika::resource::scheduling_policy scheduler)
+{
+    using ::pika::threads::scheduler_mode;
+
+    pika::init_params init_args;
+
+    init_args.cfg = {"pika.os_threads=" + std::to_string(max_threads)};
+    init_args.rp_callback = [scheduler](auto& rp, pika::program_options::variables_map const&) {
+        std::size_t pools_added = 0;
+
+        rp.set_default_pool_name("0");
+        for (pika::resource::numa_domain const& d : rp.numa_domains())
+        {
+            for (pika::resource::core const& c : d.cores())
+            {
+                for (pika::resource::pu const& p : c.pus())
+                {
+                    if (pools_added < max_threads)
+                    {
+                        std::string name = std::to_string(pools_added);
+                        rp.create_thread_pool(name, scheduler,
+                            scheduler_mode::default_mode | scheduler_mode::enable_elasticity);
+                        rp.add_resource(p, name);
+                        ++pools_added;
+                    }
+                }
+            }
+        }
+    };
+
+    pika::start(argc, argv, init_args);
+
+    for (std::size_t pool_num = 0; pool_num < max_threads; ++pool_num)
+    {
+        auto sched = ex::thread_pool_scheduler{&pika::resource::get_thread_pool(pool_num)};
+        tt::sync_wait(ex::schedule(sched) | ex::then([pool_num]() {
+            PIKA_TEST_EQ(pika::get_thread_pool_num(), pool_num);
+            PIKA_TEST_EQ(pika::get_worker_thread_num(), pool_num);
+            PIKA_TEST_EQ(pika::get_local_worker_thread_num(), static_cast<std::size_t>(0));
+        }));
+    }
+
+    pika::finalize();
+    PIKA_TEST_EQ(pika::stop(), 0);
+}
+
+int main(int argc, char* argv[])
+{
+    PIKA_ASSERT(max_threads >= 2);
+
+    std::vector<pika::resource::scheduling_policy> schedulers = {
+        pika::resource::scheduling_policy::local,
+        pika::resource::scheduling_policy::local_priority_fifo,
+#if defined(PIKA_HAVE_CXX11_STD_ATOMIC_128BIT)
+        pika::resource::scheduling_policy::local_priority_lifo,
+#endif
+#if defined(PIKA_HAVE_CXX11_STD_ATOMIC_128BIT)
+        pika::resource::scheduling_policy::abp_priority_fifo,
+        pika::resource::scheduling_policy::abp_priority_lifo,
+#endif
+        pika::resource::scheduling_policy::static_,
+        pika::resource::scheduling_policy::static_priority,
+        pika::resource::scheduling_policy::shared_priority,
+    };
+
+    for (auto const scheduler : schedulers) { test_scheduler(argc, argv, scheduler); }
+
+    return 0;
+}