use new-style sender algo customization in stream_scheduler

Author: ericniebler

.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

@@ -35,7 +35,7 @@ CompileFlags:
 
 # Use clang++ in CUDA mode to provide intellisense for files with `nvexec` in their path
 If:
-  PathMatch: (include|examples|test)/nvexec.*
+  PathMatch: .*\b(include|examples|test)/nvexec.*
 CompileFlags:
   Compiler: clang++
   Add:
@@ -54,6 +54,7 @@ CompileFlags:
     - "-ferror-limit=0"
     - "-fmacro-backtrace-limit=0"
     - "-ftemplate-backtrace-limit=0"
+    - "-std=gnu++20"
   Remove:
     - "-stdpar*"
     # strip CUDA fatbin args

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/maxwell/common.cuh

@@ -13,6 +13,9 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+
+// clang-format Language: Cpp
+
 #pragma once
 
 #include "stdexec/__detail/__config.hpp"

examples/nvexec/maxwell/cpp.cuh

@@ -13,6 +13,9 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+
+// clang-format Language: Cpp
+
 #pragma once
 
 #include "common.cuh"

examples/nvexec/maxwell/cuda.cuh

@@ -13,6 +13,9 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+
+// clang-format Language: Cpp
+
 #pragma once
 
 #include "common.cuh"

examples/nvexec/maxwell/snr.cuh

@@ -13,6 +13,9 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+
+// clang-format Language: Cpp
+
 #pragma once
 
 #include "common.cuh"
@@ -54,7 +57,7 @@ STDEXEC_PRAGMA_IGNORE_GNU("-Wmissing-braces")
 namespace ex = stdexec;
 
 #if defined(_NVHPC_CUDA) || defined(__CUDACC__)
-namespace nvexec::STDEXEC_STREAM_DETAIL_NS { //
+namespace nvexec::_strm { //
 
   namespace repeat_n {
     template <class OpT>
@@ -188,7 +191,7 @@ namespace nvexec::STDEXEC_STREAM_DETAIL_NS { //
         }});
       }
     };
-}} // namespace nvexec::STDEXEC_STREAM_DETAIL_NS::repeat_n
+}} // namespace nvexec::_strm::repeat_n
 #endif
 
 namespace repeat_n_detail {
@@ -346,7 +349,7 @@ namespace repeat_n_detail {
 #if defined(_NVHPC_CUDA) || defined(__CUDACC__)
     template <stdexec::__decays_to<repeat_n_sender_t> Self, stdexec::receiver Receiver>
       requires(stdexec::sender_to<Sender, Receiver>)
-           && (!nvexec::STDEXEC_STREAM_DETAIL_NS::receiver_with_stream_env<Receiver>)
+           && (!nvexec::_strm::receiver_with_stream_env<Receiver>)
     friend auto tag_invoke(stdexec::connect_t, Self&& self, Receiver r)
       -> repeat_n_detail::operation_state_t<SenderId, Closure, stdexec::__id<Receiver>> {
       return repeat_n_detail::operation_state_t<SenderId, Closure, stdexec::__id<Receiver>>(
@@ -355,13 +358,11 @@ namespace repeat_n_detail {
 
     template <stdexec::__decays_to<repeat_n_sender_t> Self, stdexec::receiver Receiver>
       requires(stdexec::sender_to<Sender, Receiver>)
-           && (nvexec::STDEXEC_STREAM_DETAIL_NS::receiver_with_stream_env<Receiver>)
-    friend auto
-      tag_invoke(stdexec::connect_t, Self&& self, Receiver r) -> nvexec::STDEXEC_STREAM_DETAIL_NS::
-        repeat_n::operation_state_t<SenderId, Closure, stdexec::__id<Receiver>> {
-      return nvexec::STDEXEC_STREAM_DETAIL_NS::repeat_n::
-        operation_state_t<SenderId, Closure, stdexec::__id<Receiver>>(
-          static_cast<Sender&&>(self.sender_), self.closure_, static_cast<Receiver&&>(r), self.n_);
+           && (nvexec::_strm::receiver_with_stream_env<Receiver>)
+    friend auto tag_invoke(stdexec::connect_t, Self&& self, Receiver r)
+      -> nvexec::_strm::repeat_n::operation_state_t<SenderId, Closure, stdexec::__id<Receiver>> {
+      return nvexec::_strm::repeat_n::operation_state_t<SenderId, Closure, stdexec::__id<Receiver>>(
+        static_cast<Sender&&>(self.sender_), self.closure_, static_cast<Receiver&&>(r), self.n_);
     }
 #else
     template <stdexec::__decays_to<repeat_n_sender_t> Self, stdexec::receiver Receiver>