remove code comment and add device id check

Signed-off-by: Bangtian Liu <liubangtian@gmail.com>

tuner/tuner/libtuner.py

@@ -731,10 +731,8 @@ def benchmark_candidates(candidate_indices, devices, tuning_client, candidate_tr
     """
     Runs the benchmarking for a given list of candidate indices.
     """
-    # Create worker context queue
     worker_context_queue = create_worker_context_queue(devices)
 
-    # Prepare task list
     task_list = [
         BenchmarkPack(
             iree_benchmark_module_flags=tuning_client.get_iree_benchmark_module_flags(),
@@ -745,16 +743,14 @@ def benchmark_candidates(candidate_indices, devices, tuning_client, candidate_tr
     ]
 
     # Perform benchmarking
-    benchmark_results = multiprocess_progress_wrapper(
+    return multiprocess_progress_wrapper(
         num_worker=len(devices),
         task_list=task_list,
         function=run_iree_benchmark_module_command,
         initializer=init_worker_context,
         initializer_inputs=(worker_context_queue,),
     )
 
-    return benchmark_results
-
 
 def compile(
     args: argparse.Namespace,
@@ -886,7 +882,7 @@ def get_speedup(result: BenchmarkResult) -> float:
 
     top_candidates = [result.candidate_id for result in best_results]
 
-    # Benchmarking baselines on each involved device again to check performance regression on devices.
+    # Benchmarking baselines again to check for performance regressions. These may indicate machine instability, overheating, etc.
     post_baseline_indices = [0] * len(args.devices)
     post_baseline_results = benchmark_candidates(
         candidate_indices=post_baseline_indices,
@@ -901,22 +897,22 @@ def get_speedup(result: BenchmarkResult) -> float:
 
     assert (
         baseline_times_by_device.keys() == post_baseline_times_by_device.keys()
-    ), "Error: The device IDs in baseline and post-baseline results do not match."
+    ), "Device ID mismatch between baseline runs."
 
-    regression_detected = False
     for device_id in baseline_times_by_device:
+        if device_id not in baseline_times_by_device:
+            logging.warning(f"Baseline time missing for device {device_id}")
+            continue
+
         baseline_time = baseline_times_by_device[device_id]
         post_time = post_baseline_times_by_device[device_id]
 
         if post_time > baseline_time * 1.03:
-            regression_detected = True
             percentage_slower = ((post_time - baseline_time) / baseline_time) * 100
             logging.warning(
                 f"Performance regression detected on device {device_id}: "
                 f"Baseline time = {baseline_time}, Post-baseline time = {post_time}, "
                 f"Slower by {percentage_slower:.3f}%"
             )
 
-    if not regression_detected:
-        logging.debug("No performance regressions detected.")
     return top_candidates