Merge pull request #1495 from NVIDIA/clang-tidy-and-format

apply clang-tidy and clang-format

Author: ericniebler

.clang-tidy

@@ -1,21 +1,23 @@
 ---
-Checks:          'clang-diagnostic-*,clang-analyzer-*,-*,bugprone-*,modernize-*,-bugprone-reserved-identifier,-bugprone-unchecked-optional-access,-bugprone-macro-parentheses,-bugprone-easily-swappable-parameters'
+Checks:          'clang-diagnostic-*,clang-analyzer-*,-*,bugprone-*,modernize-*,-bugprone-reserved-identifier,-bugprone-unchecked-optional-access,-bugprone-macro-parentheses,-bugprone-easily-swappable-parameters,-bugprone-dynamic-static-initializers,-modernize-avoid-c-arrays'
 WarningsAsErrors: ''
 HeaderFileExtensions:
   - ''
   - h
-  - hh
   - hpp
-  - hxx
   - cuh
 ImplementationFileExtensions:
   - c
-  - cc
   - cpp
-  - cxx
-HeaderFilterRegex: ''
+  - cu
+HeaderFilterRegex: '^.*[.](h|hpp|cuh)$'
+ExcludeHeaderFilterRegex: '^.*__system_context_default_impl_entry[.]hpp$'
 FormatStyle:     none
 User:            eniebler
+ExtraArgsBefore: [
+  '-DSTDEXEC_SYSTEM_CONTEXT_HEADER_ONLY=1',
+  '-DSTDEXEC_CLANG_TIDY_INVOKED=1',
+]
 CheckOptions:
   bugprone-argument-comment.CommentFloatLiterals: '0'
   bugprone-argument-comment.CommentUserDefinedLiterals: '0'

.clangd

@@ -52,6 +52,7 @@ CompileFlags:
     - "-fmacro-backtrace-limit=0"
     - "-ftemplate-backtrace-limit=0"
     - "-std=gnu++20"
+    - "-DSTDEXEC_CLANGD_INVOKED"
   Remove:
     - "-stdpar*"
     # strip CUDA fatbin args

examples/_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/algorithms/retry.hpp

@@ -52,11 +52,11 @@ template <class S, class R>
 struct _retry_receiver : stdexec::receiver_adaptor<_retry_receiver<S, R>> {
   _op<S, R>* o_;
 
-  R&& base() && noexcept {
+  auto base() && noexcept -> R&& {
     return static_cast<R&&>(o_->r_);
   }
 
-  const R& base() const & noexcept {
+  auto base() const & noexcept -> const R& {
     return o_->r_;
   }
 
@@ -141,6 +141,6 @@ struct _retry_sender {
 };
 
 template <stdexec::sender S>
-stdexec::sender auto retry(S s) {
+auto retry(S s) -> stdexec::sender auto {
   return _retry_sender{static_cast<S&&>(s)};
 }

examples/algorithms/then.hpp

@@ -92,6 +92,6 @@ struct _then_sender {
 };
 
 template <stdexec::sender S, class F>
-stdexec::sender auto then(S s, F f) {
+auto then(S s, F f) -> stdexec::sender auto {
   return _then_sender<S, F>{static_cast<S&&>(s), static_cast<F&&>(f)};
 }

examples/benchmark/asio_thread_pool.cpp

@@ -84,6 +84,6 @@ struct RunThread {
   }
 };
 
-int main(int argc, char** argv) {
+auto main(int argc, char** argv) -> int {
   my_main<execpools::asio_thread_pool, RunThread>(argc, argv);
 }

examples/benchmark/common.hpp

@@ -27,10 +27,11 @@
 #include <memory>
 #include <span>
 #include <thread>
+#include <utility>
 #include <vector>
 
 #if __has_include(<memory_resource>)
-#  include <memory_resource>
+#  include <memory_resource> // IWYU pragma: keep
 namespace pmr = std::pmr;
 #else
 #  define STDEXEC_NO_MONOTONIC_BUFFER_RESOURCE 1
@@ -41,15 +42,15 @@ struct statistics {
   double ops_per_sec;
 };
 
-statistics compute_perf(
+auto compute_perf(
   std::chrono::steady_clock::time_point start,
   std::chrono::steady_clock::time_point end,
-  std::size_t total_scheds) {
+  std::size_t total_scheds) -> statistics {
   auto dur = std::chrono::duration_cast<std::chrono::nanoseconds>(end - start);
   auto dur_ms = std::chrono::duration_cast<std::chrono::milliseconds>(dur);
   auto dur_dbl = std::chrono::duration_cast<std::chrono::duration<double>>(dur);
   double ops_per_sec = static_cast<double>(total_scheds) / dur_dbl.count();
-  return {dur_ms, ops_per_sec};
+  return {.total_time_ms = dur_ms, .ops_per_sec = ops_per_sec};
 }
 
 struct statistics_all {
@@ -61,12 +62,12 @@ struct statistics_all {
   double stddev;
 };
 
-statistics_all compute_perf(
+auto compute_perf(
   std::span<const std::chrono::steady_clock::time_point> start,
   std::span<const std::chrono::steady_clock::time_point> end,
   std::size_t i0,
   std::size_t i,
-  std::size_t total_scheds) {
+  std::size_t total_scheds) -> statistics_all {
   double average = 0.0;
   double max = 0.0;
   double min = std::numeric_limits<double>::max();
@@ -85,7 +86,13 @@ statistics_all compute_perf(
   variance /= static_cast<double>(i + 1 - i0);
   double stddev = std::sqrt(variance);
   auto stats = compute_perf(start[i], end[i], total_scheds);
-  statistics_all all{stats.total_time_ms, stats.ops_per_sec, average, max, min, stddev};
+  statistics_all all{
+    .total_time_ms = stats.total_time_ms,
+    .ops_per_sec = stats.ops_per_sec,
+    .average = average,
+    .max = max,
+    .min = min,
+    .stddev = stddev};
   return all;
 }
 
@@ -119,15 +126,15 @@ void my_main(int argc, char** argv, exec::numa_policy policy = exec::get_numa_po
   std::atomic<bool> stop{false};
 #ifndef STDEXEC_NO_MONOTONIC_BUFFER_RESOURCE
   std::size_t buffer_size = 2000 << 20;
-  for (std::size_t i = 0; i < static_cast<std::size_t>(nthreads); ++i) {
+  for (std::size_t i = 0; std::cmp_less(i, nthreads); ++i) {
     exec::numa_allocator<char> alloc(policy.thread_index_to_node(i));
     buffers.push_back(
       std::unique_ptr<char, numa_deleter>{
-        alloc.allocate(buffer_size), numa_deleter{buffer_size, alloc}
+        alloc.allocate(buffer_size), numa_deleter{.size_ = buffer_size, .allocator_ = alloc}
     });
   }
 #endif
-  for (std::size_t i = 0; i < static_cast<std::size_t>(nthreads); ++i) {
+  for (std::size_t i = 0; std::cmp_less(i, nthreads); ++i) {
     threads.emplace_back(
       RunThread{},
       std::ref(*pool),

examples/benchmark/fibonacci.cpp

@@ -23,7 +23,7 @@
 #include <exec/any_sender_of.hpp>
 #include <stdexec/execution.hpp>
 
-long serial_fib(long n) {
+auto serial_fib(long n) -> long {
   return n < 2 ? n : serial_fib(n - 1) + serial_fib(n - 2);
 }
 
@@ -67,7 +67,7 @@ struct fib_s {
   };
 
   template <stdexec::receiver_of<completion_signatures> Receiver>
-  friend operation<Receiver> tag_invoke(stdexec::connect_t, fib_s self, Receiver rcvr) {
+  friend auto tag_invoke(stdexec::connect_t, fib_s self, Receiver rcvr) -> operation<Receiver> {
     return {static_cast<Receiver&&>(rcvr), self.cutoff, self.n, self.sched};
   }
 };
@@ -82,7 +82,7 @@ auto measure(F&& f) {
   return std::chrono::duration_cast<duration>(std::chrono::steady_clock::now() - start).count();
 }
 
-int main(int argc, char** argv) {
+auto main(int argc, char** argv) -> int {
   if (argc < 5) {
     std::cerr << "Usage: example.benchmark.fibonacci cutoff n nruns {tbb|static}" << std::endl;
     return -1;

examples/benchmark/static_thread_pool.cpp

@@ -87,12 +87,13 @@ struct RunThread {
 };
 
 struct my_numa_distribution : public exec::default_numa_policy {
-  int thread_index_to_node(std::size_t index) const noexcept {
+  [[nodiscard]]
+  auto thread_index_to_node(std::size_t index) const noexcept -> int {
     return exec::default_numa_policy::thread_index_to_node(2 * index);
   }
 };
 
-int main(int argc, char** argv) {
+auto main(int argc, char** argv) -> int {
   my_numa_distribution numa{};
   my_main<exec::static_thread_pool, RunThread>(argc, argv, numa);
 }

examples/benchmark/static_thread_pool_bulk_enqueue.cpp

@@ -58,12 +58,13 @@ struct RunThread {
 };
 
 struct my_numa_distribution : public exec::default_numa_policy {
-  int thread_index_to_node(std::size_t index) const noexcept {
+  [[nodiscard]]
+  auto thread_index_to_node(std::size_t index) const noexcept -> int {
     return exec::default_numa_policy::thread_index_to_node(2 * index);
   }
 };
 
-int main(int argc, char** argv) {
+auto main(int argc, char** argv) -> int {
   exec::numa_policy numa{my_numa_distribution{}};
   my_main<exec::static_thread_pool, RunThread>(argc, argv, std::move(numa));
 }