fix two bugs related to overhead

We were always dividing the total function execution time by
the number of cores, however sometimes the user sets 'ntasks'
so instead we should divide by '_get_max_tasks()'.

Also the future start_time wasn't measured in the case where
one uses a SequentialExecutor because the Future that is
returned already contains the result.

Author: basnijholt

adaptive/runner.py

@@ -201,7 +201,7 @@ def _process_futures(self, done_futs):
                     if self.raise_if_retries_exceeded:
                         self._do_raise(e, x)
             else:
-                self._elapsed_function_time += t / _get_ncores(self.executor)
+                self._elapsed_function_time += t / self._get_max_tasks()
                 self.to_retry.pop(x, None)
                 self.tracebacks.pop(x, None)
                 if self.do_log:
@@ -220,8 +220,9 @@ def _get_futures(self):
         points, _ = self._ask(n_new_tasks)
 
         for x in points:
+            start_time = time.time()  # so we can measure execution time
             fut = self._submit(x)
-            fut.start_time = time.time()  # so we can measure execution time
+            fut.start_time = start_time
             self.pending_points[fut] = x
 
         # Collect and results and add them to the learner