[Flang][runtime] Distinguish CPU time and elapsed time for cpu_time and system_clock

flang/runtime/time-intrinsic.cpp

@@ -19,6 +19,7 @@
 #include <cstdlib>
 #include <cstring>
 #include <ctime>
+#include <chrono>
 #ifdef _WIN32
 #include "flang/Common/windows-include.h"
 #else
@@ -64,20 +65,29 @@ template <typename Unused = void> double GetCpuTime(fallback_implementation) {
 // clock_gettime is implemented in the pthread library for MinGW.
 // Using it here would mean that all programs that link libFortranRuntime are
 // required to also link to pthread. Instead, don't use the function.
-#undef CLOCKID
-#elif defined CLOCK_PROCESS_CPUTIME_ID
-#define CLOCKID CLOCK_PROCESS_CPUTIME_ID
+#undef CLOCKID_CPU_TIME
+#undef CLOCKID_ELAPSED_TIME
+#else
+// Determine what clock to use for CPU time.
+#if defined CLOCK_PROCESS_CPUTIME_ID
+#define CLOCKID_CPU_TIME CLOCK_PROCESS_CPUTIME_ID
 #elif defined CLOCK_THREAD_CPUTIME_ID
-#define CLOCKID CLOCK_THREAD_CPUTIME_ID
-#elif defined CLOCK_MONOTONIC
-#define CLOCKID CLOCK_MONOTONIC
+#define CLOCKID_CPU_TIME CLOCK_THREAD_CPUTIME_ID
+#else
+#undef CLOCKID_CPU_TIME
+#endif
+
+// Determine what clock to use for elapsed time.
+#if defined CLOCK_MONOTONIC
+#define CLOCKID_ELAPSED_TIME CLOCK_MONOTONIC
 #elif defined CLOCK_REALTIME
-#define CLOCKID CLOCK_REALTIME
+#define CLOCKID_ELAPSED_TIME CLOCK_REALTIME
 #else
-#undef CLOCKID
+#undef CLOCKID_ELAPSED_TIME
+#endif
 #endif
 
-#ifdef CLOCKID
+#ifdef CLOCKID_CPU_TIME
 // POSIX implementation using clock_gettime. This is only enabled where
 // clock_gettime is available.
 template <typename T = int, typename U = struct timespec>
@@ -86,13 +96,13 @@ double GetCpuTime(preferred_implementation,
     T ClockId = 0, U *Timespec = nullptr,
     decltype(clock_gettime(ClockId, Timespec)) *Enabled = nullptr) {
   struct timespec tspec;
-  if (clock_gettime(CLOCKID, &tspec) == 0) {
+  if (clock_gettime(CLOCKID_CPU_TIME, &tspec) == 0) {
     return tspec.tv_nsec * 1.0e-9 + tspec.tv_sec;
   }
   // Return some negative value to represent failure.
   return -1.0;
 }
-#endif
+#endif // CLOCKID_CPU_TIME
 
 using count_t = std::int64_t;
 using unsigned_count_t = std::uint64_t;
@@ -105,31 +115,24 @@ static constexpr inline unsigned_count_t GetHUGE(int kind) {
   return (unsigned_count_t{1} << ((8 * kind) - 1)) - 1;
 }
 
-// This is the fallback implementation, which should work everywhere. Note that
-// in general we can't recover after std::clock has reached its maximum value.
+// This is the fallback implementation, which should work everywhere.
 template <typename Unused = void>
 count_t GetSystemClockCount(int kind, fallback_implementation) {
-  std::clock_t timestamp{std::clock()};
-  if (timestamp == static_cast<std::clock_t>(-1)) {
+  unsigned_count_t timestamp;
+  timestamp =
+      std::chrono::high_resolution_clock::now().time_since_epoch().count();
+  if (timestamp == static_cast<unsigned_count_t>(-1)) {
     // Return -HUGE(COUNT) to represent failure.
     return -static_cast<count_t>(GetHUGE(kind));
   }
-  // Convert the timestamp to std::uint64_t with wrap-around. The timestamp is
-  // most likely a floating-point value (since C'11), so compute the modulus
-  // carefully when one is required.
-  constexpr auto maxUnsignedCount{std::numeric_limits<unsigned_count_t>::max()};
-  if constexpr (std::numeric_limits<std::clock_t>::max() > maxUnsignedCount) {
-    timestamp -= maxUnsignedCount * std::floor(timestamp / maxUnsignedCount);
-  }
-  unsigned_count_t unsignedCount{static_cast<unsigned_count_t>(timestamp)};
   // Return the modulus of the unsigned integral count with HUGE(COUNT)+1.
   // The result is a signed integer but never negative.
-  return static_cast<count_t>(unsignedCount % (GetHUGE(kind) + 1));
+  return static_cast<count_t>(timestamp % (GetHUGE(kind) + 1));
 }
 
 template <typename Unused = void>
 count_t GetSystemClockCountRate(int kind, fallback_implementation) {
-  return CLOCKS_PER_SEC;
+  return std::chrono::high_resolution_clock::period::den;
 }
 
 template <typename Unused = void>
@@ -149,15 +152,15 @@ constexpr unsigned_count_t DS_PER_SEC{10u};
 constexpr unsigned_count_t MS_PER_SEC{1'000u};
 constexpr unsigned_count_t NS_PER_SEC{1'000'000'000u};
 
-#ifdef CLOCKID
+#ifdef CLOCKID_ELAPSED_TIME
 template <typename T = int, typename U = struct timespec>
 count_t GetSystemClockCount(int kind, preferred_implementation,
     // We need some dummy parameters to pass to decltype(clock_gettime).
     T ClockId = 0, U *Timespec = nullptr,
     decltype(clock_gettime(ClockId, Timespec)) *Enabled = nullptr) {
   struct timespec tspec;
   const unsigned_count_t huge{GetHUGE(kind)};
-  if (clock_gettime(CLOCKID, &tspec) != 0) {
+  if (clock_gettime(CLOCKID_ELAPSED_TIME, &tspec) != 0) {
     return -huge; // failure
   }
   unsigned_count_t sec{static_cast<unsigned_count_t>(tspec.tv_sec)};
@@ -170,7 +173,7 @@ count_t GetSystemClockCount(int kind, preferred_implementation,
     return (sec * DS_PER_SEC + (nsec / (NS_PER_SEC / DS_PER_SEC))) % (huge + 1);
   }
 }
-#endif
+#endif // CLOCKID_ELAPSED_TIME
 
 template <typename T = int, typename U = struct timespec>
 count_t GetSystemClockCountRate(int kind, preferred_implementation,

flang/test/Runtime/no-cpp-dep.c

@@ -30,12 +30,14 @@ int32_t RTNAME(ArgumentCount)();
 int32_t RTNAME(GetCommandArgument)(int32_t, const struct Descriptor *,
     const struct Descriptor *, const struct Descriptor *);
 int32_t RTNAME(GetEnvVariable)();
+int64_t RTNAME(SystemClockCount)(int kind);
 
 int main() {
   double x = RTNAME(CpuTime)();
   RTNAME(ProgramStart)(0, 0, 0, 0);
   int32_t c = RTNAME(ArgumentCount)();
   int32_t v = RTNAME(GetCommandArgument)(0, 0, 0, 0);
   int32_t e = RTNAME(GetEnvVariable)("FOO", 0, 0);
+  int64_t t = RTNAME(SystemClockCount)(8);
   return x + c + v + e;
 }

flang/test/Runtime/system_clock.f90

@@ -0,0 +1,27 @@
+! RUN: %flang -o %t %s
+! RUN: %t
+
+program system_clock_test
+    use iso_fortran_env, only: int64, real64
+    implicit none
+
+    integer, parameter :: delta = 1
+    real(kind=real64), parameter :: epsilon = 0.001
+
+    integer(kind=int64) :: t_start, t_end
+    integer(kind=int64) :: rate
+    real(kind=real64) :: diff
+
+    call system_clock(count_rate=rate)
+
+    call system_clock(t_start)
+    call sleep(delta)
+    call system_clock(t_end)
+
+    diff = real(t_end - t_start, kind=real64) / real(rate, kind=real64)
+
+    if (abs(diff - real(delta, kind=real64)) <= epsilon) then
+        stop 0, quiet=.true.
+    end if
+    stop 1, quiet=.true.
+end program system_clock_test