Implement speedup stack graph generation

benchkit/benchmark.py

@@ -30,14 +30,7 @@
 )
 from benchkit.utils.system import get_boot_args
 from benchkit.utils.tee import teeprint
-from benchkit.utils.types import (
-    Command,
-    Constants,
-    Environment,
-    PathType,
-    Pretty,
-    SplitCommand,
-)
+from benchkit.utils.types import Command, Constants, Environment, PathType, Pretty, SplitCommand
 from benchkit.utils.variables import list_groupby
 
 RecordKey = str

benchkit/cli/init.py

@@ -7,11 +7,7 @@
 import black
 import isort
 
-from benchkit.cli.generate import (
-    generate_benchmark,
-    generate_campaign,
-    get_gitignore_content,
-)
+from benchkit.cli.generate import generate_benchmark, generate_campaign, get_gitignore_content
 from benchkit.utils.misc import get_benchkit_temp_folder_str
 
 _DOTGIT_DIR = Path(".git")

benchkit/commandattachments/offcputime.py

@@ -46,6 +46,7 @@ class Offcputime(LibbpfTools):
                         (default 1024)
         state: filter on this thread state bitmask (eg, 2 == TASK_UNINTERRUPTIBLE)
                         see include/linux/sched.h
+        filter_comm: filter out the given comm names
     """
 
     def __init__(
@@ -60,6 +61,7 @@ def __init__(
         perf_max_stack_depth: int = -1,
         stack_storage_size: int = -1,
         state: int = -1,
+        filter_comm: list[str] = [],
         platform: Platform = None,
     ) -> None:
 
@@ -78,6 +80,7 @@ def __init__(
         self._perf_max_stack_depth = perf_max_stack_depth
         self._stack_storage_size = stack_storage_size
         self._state = state
+        self._filter_comm = filter_comm
 
         self.out_file_name = "offcputime.out"
         self.err_file_name = "offcputime.err"
@@ -166,6 +169,9 @@ def post_run_hook(
                 m = row_re.search(line)
                 if m:
                     name = m.group(1)
+                    if name in self._filter_comm:
+                        continue
+
                     pid = int(m.group(2))
                     delta_micro_s = int(m.group(3))
 
@@ -190,6 +196,9 @@ def post_run_hook(
                 (sum(d["total_off_time_micro_s"] for d in per_pid_dict.values()) / number_of_pids)
                 if number_of_pids != 0
                 else 0
-            )
+            ),
+            "offcputime_total_micro_s": sum(
+                d["total_off_time_micro_s"] for d in per_pid_dict.values()
+            ),
         }
         return return_dict