python
diff --git a/‎Doc/library/time.rst‎
Lines changed: 27 additions & 0 deletions b/‎Doc/library/time.rst‎
Lines changed: 27 additions & 0 deletions
diff --git a/‎Doc/whatsnew/3.7.rst‎
Lines changed: 4 additions & 0 deletions b/‎Doc/whatsnew/3.7.rst‎
Lines changed: 4 additions & 0 deletions
diff --git a/‎Lib/test/test_time.py‎
Lines changed: 57 additions & 0 deletions b/‎Lib/test/test_time.py‎
Lines changed: 57 additions & 0 deletions
diff --git a/‎Misc/NEWS.d/next/Library/2017-11-15-20-03-45.bpo-32025.lnIKYT.rst‎
Lines changed: 1 addition & 0 deletions b/‎Misc/NEWS.d/next/Library/2017-11-15-20-03-45.bpo-32025.lnIKYT.rst‎
Lines changed: 1 addition & 0 deletions
diff --git a/‎Modules/timemodule.c‎
Lines changed: 117 additions & 0 deletions b/‎Modules/timemodule.c‎
Lines changed: 117 additions & 0 deletions
@@ -241,6 +241,7 @@ Functions
    * ``'monotonic'``: :func:`time.monotonic`
    * ``'perf_counter'``: :func:`time.perf_counter`
    * ``'process_time'``: :func:`time.process_time`
+   * ``'thread_time'``: :func:`time.thread_time`
    * ``'time'``: :func:`time.time`
 
    The result has the following attributes:
@@ -603,6 +604,32 @@ Functions
    of the calendar date may be accessed as attributes.
 
 
+.. function:: thread_time() -> float
+
+   .. index::
+      single: CPU time
+      single: processor time
+      single: benchmarking
+
+   Return the value (in fractional seconds) of the sum of the system and user
+   CPU time of the current thread.  It does not include time elapsed during
+   sleep.  It is thread-specific by definition.  The reference point of the
+   returned value is undefined, so that only the difference between the results
+   of consecutive calls in the same thread is valid.
+
+   Availability:  Windows, Linux, Unix systems supporting
+   ``CLOCK_THREAD_CPUTIME_ID``.
+
+   .. versionadded:: 3.7
+
+
+.. function:: thread_time_ns() -> int
+
+   Similar to :func:`thread_time` but return time as nanoseconds.
+
+   .. versionadded:: 3.7
+
+
 .. function:: time_ns() -> int
 
    Similar to :func:`time` but returns time as an integer number of nanoseconds
 
@@ -372,6 +372,10 @@ Add new clock identifiers:
   the time the system has been running and not suspended, providing accurate
   uptime measurement, both absolute and interval.
 
+Added functions :func:`time.thread_time` and :func:`time.thread_time_ns`
+to get per-thread CPU time measurements.
+(Contributed by Antoine Pitrou in :issue:`32025`.)
+
 unittest.mock
 -------------
 
 
@@ -47,6 +47,12 @@ class _PyTime(enum.IntEnum):
 )
 
 
+def busy_wait(duration):
+    deadline = time.monotonic() + duration
+    while time.monotonic() < deadline:
+        pass
+
+
 class TimeTestCase(unittest.TestCase):
 
     def setUp(self):
@@ -81,6 +87,10 @@ def check_ns(sec, ns):
         check_ns(time.process_time(),
                  time.process_time_ns())
 
+        if hasattr(time, 'thread_time'):
+            check_ns(time.thread_time(),
+                     time.thread_time_ns())
+
         if hasattr(time, 'clock_gettime'):
             check_ns(time.clock_gettime(time.CLOCK_REALTIME),
                      time.clock_gettime_ns(time.CLOCK_REALTIME))
@@ -486,10 +496,57 @@ def test_process_time(self):
         # on Windows
         self.assertLess(stop - start, 0.020)
 
+        # process_time() should include CPU time spent in any thread
+        start = time.process_time()
+        busy_wait(0.100)
+        stop = time.process_time()
+        self.assertGreaterEqual(stop - start, 0.020)  # machine busy?
+
+        t = threading.Thread(target=busy_wait, args=(0.100,))
+        start = time.process_time()
+        t.start()
+        t.join()
+        stop = time.process_time()
+        self.assertGreaterEqual(stop - start, 0.020)  # machine busy?
+
         info = time.get_clock_info('process_time')
         self.assertTrue(info.monotonic)
         self.assertFalse(info.adjustable)
 
+    def test_thread_time(self):
+        if not hasattr(time, 'thread_time'):
+            if sys.platform.startswith(('linux', 'win')):
+                self.fail("time.thread_time() should be available on %r"
+                          % (sys.platform,))
+            else:
+                self.skipTest("need time.thread_time")
+
+        # thread_time() should not include time spend during a sleep
+        start = time.thread_time()
+        time.sleep(0.100)
+        stop = time.thread_time()
+        # use 20 ms because thread_time() has usually a resolution of 15 ms
+        # on Windows
+        self.assertLess(stop - start, 0.020)
+
+        # thread_time() should include CPU time spent in current thread...
+        start = time.thread_time()
+        busy_wait(0.100)
+        stop = time.thread_time()
+        self.assertGreaterEqual(stop - start, 0.020)  # machine busy?
+
+        # ...but not in other threads
+        t = threading.Thread(target=busy_wait, args=(0.100,))
+        start = time.thread_time()
+        t.start()
+        t.join()
+        stop = time.thread_time()
+        self.assertLess(stop - start, 0.020)
+
+        info = time.get_clock_info('thread_time')
+        self.assertTrue(info.monotonic)
+        self.assertFalse(info.adjustable)
+
     @unittest.skipUnless(hasattr(time, 'clock_settime'),
                          'need time.clock_settime')
     def test_monotonic_settime(self):
 
@@ -0,0 +1 @@
+Add time.thread_time() and time.thread_time_ns()
@@ -1258,6 +1258,112 @@ Process time for profiling as nanoseconds:\n\
 sum of the kernel and user-space CPU time.");
 
 
+#if defined(MS_WINDOWS)
+#define HAVE_THREAD_TIME
+static int
+_PyTime_GetThreadTimeWithInfo(_PyTime_t *tp, _Py_clock_info_t *info)
+{
+    HANDLE thread;
+    FILETIME creation_time, exit_time, kernel_time, user_time;
+    ULARGE_INTEGER large;
+    _PyTime_t ktime, utime, t;
+    BOOL ok;
+
+    thread =  GetCurrentThread();
+    ok = GetThreadTimes(thread, &creation_time, &exit_time,
+                        &kernel_time, &user_time);
+    if (!ok) {
+        PyErr_SetFromWindowsErr(0);
+        return -1;
+    }
+
+    if (info) {
+        info->implementation = "GetThreadTimes()";
+        info->resolution = 1e-7;
+        info->monotonic = 1;
+        info->adjustable = 0;
+    }
+
+    large.u.LowPart = kernel_time.dwLowDateTime;
+    large.u.HighPart = kernel_time.dwHighDateTime;
+    ktime = large.QuadPart;
+
+    large.u.LowPart = user_time.dwLowDateTime;
+    large.u.HighPart = user_time.dwHighDateTime;
+    utime = large.QuadPart;
+
+    /* ktime and utime have a resolution of 100 nanoseconds */
+    t = _PyTime_FromNanoseconds((ktime + utime) * 100);
+    *tp = t;
+    return 0;
+}
+
+#elif defined(HAVE_CLOCK_GETTIME) && defined(CLOCK_PROCESS_CPUTIME_ID)
+#define HAVE_THREAD_TIME
+static int
+_PyTime_GetThreadTimeWithInfo(_PyTime_t *tp, _Py_clock_info_t *info)
+{
+    struct timespec ts;
+    const clockid_t clk_id = CLOCK_THREAD_CPUTIME_ID;
+    const char *function = "clock_gettime(CLOCK_THREAD_CPUTIME_ID)";
+
+    if (clock_gettime(clk_id, &ts)) {
+        PyErr_SetFromErrno(PyExc_OSError);
+        return -1;
+    }
+    if (info) {
+        struct timespec res;
+        info->implementation = function;
+        info->monotonic = 1;
+        info->adjustable = 0;
+        if (clock_getres(clk_id, &res)) {
+            PyErr_SetFromErrno(PyExc_OSError);
+            return -1;
+        }
+        info->resolution = res.tv_sec + res.tv_nsec * 1e-9;
+    }
+
+    if (_PyTime_FromTimespec(tp, &ts) < 0) {
+        return -1;
+    }
+    return 0;
+}
+#endif
+
+#ifdef HAVE_THREAD_TIME
+static PyObject *
+time_thread_time(PyObject *self, PyObject *unused)
+{
+    _PyTime_t t;
+    if (_PyTime_GetThreadTimeWithInfo(&t, NULL) < 0) {
+        return NULL;
+    }
+    return _PyFloat_FromPyTime(t);
+}
+
+PyDoc_STRVAR(thread_time_doc,
+"thread_time() -> float\n\
+\n\
+Thread time for profiling: sum of the kernel and user-space CPU time.");
+
+static PyObject *
+time_thread_time_ns(PyObject *self, PyObject *unused)
+{
+    _PyTime_t t;
+    if (_PyTime_GetThreadTimeWithInfo(&t, NULL) < 0) {
+        return NULL;
+    }
+    return _PyTime_AsNanosecondsObject(t);
+}
+
+PyDoc_STRVAR(thread_time_ns_doc,
+"thread_time() -> int\n\
+\n\
+Thread time for profiling as nanoseconds:\n\
+sum of the kernel and user-space CPU time.");
+#endif
+
+
 static PyObject *
 time_get_clock_info(PyObject *self, PyObject *args)
 {
@@ -1311,6 +1417,13 @@ time_get_clock_info(PyObject *self, PyObject *args)
             return NULL;
         }
     }
+#ifdef HAVE_THREAD_TIME
+    else if (strcmp(name, "thread_time") == 0) {
+        if (_PyTime_GetThreadTimeWithInfo(&t, &info) < 0) {
+            return NULL;
+        }
+    }
+#endif
     else {
         PyErr_SetString(PyExc_ValueError, "unknown clock");
         return NULL;
@@ -1519,6 +1632,10 @@ static PyMethodDef time_methods[] = {
     {"monotonic_ns",    time_monotonic_ns, METH_NOARGS, monotonic_ns_doc},
     {"process_time",    time_process_time, METH_NOARGS, process_time_doc},
     {"process_time_ns", time_process_time_ns, METH_NOARGS, process_time_ns_doc},
+#ifdef HAVE_THREAD_TIME
+    {"thread_time",     time_thread_time, METH_NOARGS, thread_time_doc},
+    {"thread_time_ns",  time_thread_time_ns, METH_NOARGS, thread_time_ns_doc},
+#endif
     {"perf_counter",    time_perf_counter, METH_NOARGS, perf_counter_doc},
     {"perf_counter_ns", time_perf_counter_ns, METH_NOARGS, perf_counter_ns_doc},
     {"get_clock_info",  time_get_clock_info, METH_VARARGS, get_clock_info_doc},
Original file line number	Diff line number	Diff line change
`@@ -0,0 +1 @@`
	`1`	`+Add time.thread_time() and time.thread_time_ns()`