Merge remote-tracking branch 'upstream/main' into syssched-R6-work

Author: lucteo

.clang-format

@@ -83,14 +83,12 @@ KeepEmptyLinesAtTheStartOfBlocks: true
 LambdaBodyIndentation: Signature
 LineEnding: LF
 Macros: [
-  'STDEXEC_MEMFN_DECL(X)=X',
-  'STDEXEC_MEMFN_DECL(X,Y)=X,Y',
-  'STDEXEC_MEMFN_DECL(X,Y,Z)=X,Y,Z',
+  'STDEXEC_MEMFN_DECL(...)=__VA_ARGS__',
   'STDEXEC_ATTRIBUTE(X)=[[]]',
-  'STDEXEC_NO_UNIQUE_ADDRESS=[[no_unique_address]]',
   'STDEXEC_IMMOVABLE_NO_UNIQUE_ADDRESS=[[no_unique_address]]',
   'STDEXEC_MISSING_MEMBER(X,Y)=true',
-  'STDEXEC_DEFINE_MEMBER(X)=void foo() {}'
+  'STDEXEC_DEFINE_MEMBER(X)=void foo() {}',
+  'STDEXEC_AUTO_RETURN(...)=->decltype(auto){ return __VA_ARGS__; }',
 ]
 MaxEmptyLinesToKeep: 2
 NamespaceIndentation: All

.clang-tidy

@@ -7,6 +7,7 @@ HeaderFileExtensions:
   - hh
   - hpp
   - hxx
+  - cuh
 ImplementationFileExtensions:
   - c
   - cc

.clangd

@@ -1,11 +1,5 @@
 # https://clangd.llvm.org/config
 
-# Apply a config conditionally to all C files
-If:
-  PathMatch: .*\.(c|h)$
-
----
-
 # Apply a config conditionally to all C++ files except those in the
 # include/execpools directory which need platform-dependent headers
 If:
@@ -17,6 +11,7 @@ If:
 # Apply a config conditionally to all CUDA files
 If:
   PathMatch: .*\.cuh?$
+  PathExclude: examples/nvexec/.*
 CompileFlags:
   Add:
     # Allow variadic CUDA functions
@@ -35,25 +30,28 @@ CompileFlags:
 
 # Use clang++ in CUDA mode to provide intellisense for files with `nvexec` in their path
 If:
-  PathMatch: (include|examples|test)/nvexec.*
+  PathMatch: [include/nvexec/.*, test/nvexec/.*]
 CompileFlags:
   Compiler: clang++
   Add:
     - "-x"
     - "cuda"
     - "-std=gnu++20"
     - "-Wno-unknown-cuda-version"
+    - "--cuda-host-only"
 
 ---
 
 # Tweak the clangd parse settings for all files
 CompileFlags:
   CompilationDatabase: .
   Add:
+    - "-DLIBCUDACXX_ENABLE_EXPERIMENTAL_MEMORY_RESOURCE"
     # report all errors
     - "-ferror-limit=0"
     - "-fmacro-backtrace-limit=0"
     - "-ftemplate-backtrace-limit=0"
+    - "-std=gnu++20"
   Remove:
     - "-stdpar*"
     # strip CUDA fatbin args
@@ -74,6 +72,7 @@ CompileFlags:
     - "--expt-relaxed-constexpr"
     - "-forward-unknown-to-host-compiler"
     - "-Werror=cross-execution-space-call"
+    - "-e1000"
 Diagnostics:
   Suppress:
     - "variadic_device_fn"

.vscode/launch.json

@@ -5,61 +5,61 @@
     "version": "0.2.0",
     "configurations": [
         {
-            "name": "C++ Tests (lldb)",
-            "type": "lldb",
+            "name": "C++ Current Target (lldb-dap)",
+            "type": "lldb-dap",
             "request": "launch",
-            "stdio": null,
             "stopOnEntry": false,
-            "terminal": "console",
-            "console": "internalConsole",
-            "sourceLanguages": ["cpp", "cuda"],
-            "internalConsoleOptions": "neverOpen",
+            "internalConsoleOptions": "openOnFirstSessionStart",
+            "cwd": "${command:cmake.launchTargetDirectory}",
+            "program": "${command:cmake.launchTargetPath}",
+            "initCommands": [
+                "settings set target.disable-aslr false"
+            ],
+            "args": "${input:CXX_PROGRAM_ARGS}",
+        },
+        {
+            "name": "C++ Tests (lldb-dap)",
+            "type": "lldb-dap",
+            "request": "launch",
+            "stopOnEntry": false,
+            "internalConsoleOptions": "openOnFirstSessionStart",
             "cwd": "${command:cmake.buildDirectory}",
-            "relativePathBase": "${command:cmake.buildDirectory}",
             "program": "${command:cmake.buildDirectory}/${input:CXX_TEST_SUITE}",
-            "initCommands": ["settings set target.disable-aslr false"],
-            "args": ["-v", "normal", "${input:CXX_TEST_TAGS}"],
+            "initCommands": [
+                "settings set target.disable-aslr false"
+            ],
+            "args": [
+                "-v",
+                "normal",
+                "${input:CXX_TEST_TAGS}"
+            ],
         },
         {
-            "name": "CUDA Tests (cuda-gdb)",
+            "name": "CUDA Current Target (cuda-gdb)",
             "type": "cuda-gdb",
             "request": "launch",
             "stopAtEntry": false,
             "breakOnLaunch": false,
-            "internalConsoleOptions": "neverOpen",
-            "program": "${command:cmake.buildDirectory}/${input:CUDA_TEST_SUITE}",
-            "cwd": "${command:cmake.buildDirectory}",
-            "args": "-v normal ${input:CUDA_TEST_TAGS}",
-        },
-        {
-            "name": "C++ Current Target (lldb)",
-            "type": "lldb",
-            "request": "launch",
-            "stdio": null,
-            "stopOnEntry": false,
-            "terminal": "console",
-            "console": "internalConsole",
-            "sourceLanguages": ["cpp", "cuda"],
-            "internalConsoleOptions": "neverOpen",
-            "cwd": "${command:cmake.launchTargetDirectory}",
-            "relativePathBase": "${command:cmake.launchTargetDirectory}",
+            "internalConsoleOptions": "openOnFirstSessionStart",
             "program": "${command:cmake.launchTargetPath}",
-            "initCommands": ["settings set target.disable-aslr false"],
+            "cwd": "${command:cmake.launchTargetDirectory}",
             "args": "${input:CXX_PROGRAM_ARGS}",
         },
         {
-            "name": "CUDA Current Target (cuda-gdb)",
+            "name": "CUDA Tests (cuda-gdb)",
             "type": "cuda-gdb",
             "request": "launch",
             "stopAtEntry": false,
             "breakOnLaunch": false,
-            "internalConsoleOptions": "neverOpen",
-            "program": "${command:cmake.launchTargetPath}",
-            "cwd": "${command:cmake.launchTargetDirectory}",
-            "args": "${input:CXX_PROGRAM_ARGS}",
+            "internalConsoleOptions": "openOnSessionStart",
+            "program": "${command:cmake.buildDirectory}/${input:CUDA_TEST_SUITE}",
+            "cwd": "${command:cmake.buildDirectory}",
+            "args": "-v normal ${input:CUDA_TEST_TAGS}",
         },
     ],
     "inputs": [
+        // These require the Tasks Shell Input extension:
+        // https://marketplace.visualstudio.com/items?itemName=augustocdias.tasks-shell-input
         {
             "id": "CXX_PROGRAM_ARGS",
             "type": "promptString",
@@ -109,4 +109,4 @@
             }
         },
     ],
-}
+}

examples/benchmark/static_thread_pool_bulk_enqueue.cpp

@@ -45,7 +45,7 @@ struct RunThread {
       auto env = exec::make_env(stdexec::prop{stdexec::get_allocator, alloc});
       auto [start, end] = exec::_pool_::even_share(total_scheds, tid, pool.available_parallelism());
       auto iterate = exec::schedule_all(pool, std::views::iota(start, end))
-                   | exec::ignore_all_values() | exec::write(env);
+                   | exec::ignore_all_values() | exec::write_env(env);
 #  else
       auto [start, end] = exec::_pool_::even_share(total_scheds, tid, pool.available_parallelism());
       auto iterate = exec::schedule_all(pool, std::views::iota(start, end))

examples/benchmark/static_thread_pool_bulk_enqueue_nested.cpp

@@ -46,7 +46,7 @@ struct RunThread {
       auto env = exec::make_env(stdexec::prop{stdexec::get_allocator, alloc});
       auto [start, end] = exec::_pool_::even_share(total_scheds, tid, pool.available_parallelism());
       auto iterate = exec::iterate(std::views::iota(start, end)) | exec::ignore_all_values()
-                   | exec::write(env);
+                   | exec::write_env(env);
 #  else
       auto [start, end] = exec::_pool_::even_share(total_scheds, tid, pool.available_parallelism());
       auto iterate = exec::iterate(std::views::iota(start, end)) | exec::ignore_all_values();

examples/nvexec/CMakeLists.txt

@@ -92,6 +92,9 @@ function(def_cpu_example example)
                 stdexec_executable_flags
     )
     set_source_files_properties(${source} PROPERTIES LANGUAGE ${_lang_cxx})
+    if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+        target_link_options(${target} PRIVATE -lc++abi -ltbb)
+    endif()
 endfunction()
 
 foreach(example ${nvexec_cpu_examples})

examples/nvexec/_clangd_helper_file.cpp

@@ -28,5 +28,5 @@
 // than those in the parent (for example, if the subdirectory has a CMakeList.txt that defines additional executables).
 // This ensures clangd provides useful intellisense for headers in any subdirectory with a CMakeList.txt.
 
-int main(void) {
+auto main() -> int {
 }

examples/nvexec/bulk.cpp

@@ -21,7 +21,7 @@
 
 namespace ex = stdexec;
 
-int main() {
+auto main() -> int {
   using nvexec::is_on_gpu;
 
   nvexec::stream_context stream_ctx{};

examples/nvexec/launch.cpp

@@ -14,6 +14,7 @@
  * limitations under the License.
  */
 
+#include <algorithm>
 #include <nvexec/stream_context.cuh>
 #include <stdexec/execution.hpp>
 
@@ -26,16 +27,18 @@ constexpr std::size_t N = 2 * 1024;
 constexpr std::size_t THREAD_BLOCK_SIZE = 128u;
 constexpr std::size_t NUM_BLOCKS = (N + THREAD_BLOCK_SIZE - 1) / THREAD_BLOCK_SIZE;
 
-#define scaling 2
+enum {
+  scaling = 2
+};
 
-int bench() {
+auto bench() -> int {
   std::vector<int> input(N, 0);
   std::iota(input.begin(), input.end(), 1);
-  std::transform(input.begin(), input.end(), input.begin(), [](int i) { return i * scaling; });
+  std::ranges::transform(input, input.begin(), [](int i) { return i * scaling; });
   return std::accumulate(input.begin(), input.end(), 0);
 }
 
-int main() {
+auto main() -> int {
   thrust::device_vector<int> input(N, 0);
   std::iota(input.begin(), input.end(), 1);
   int* first = thrust::raw_pointer_cast(input.data());
@@ -45,7 +48,7 @@ int main() {
 
   auto snd = stdexec::transfer_just(stream.get_scheduler(), first, last)
            | nvexec::launch(
-               {NUM_BLOCKS, THREAD_BLOCK_SIZE},
+               {.grid_size = NUM_BLOCKS, .block_size = THREAD_BLOCK_SIZE},
                [](cudaStream_t stm, int* first, int* last) {
                  assert(nvexec::is_on_gpu());
                  int32_t idx = blockIdx.x * blockDim.x + threadIdx.x;