Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(773)

Unified Diff: services/blamer/heap_objects/blame_data.h

Issue 2885363004: [Hacky prototype] Create a shared-memory high-performance reporting service.
Patch Set: Created 3 years, 7 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
Index: services/blamer/heap_objects/blame_data.h
diff --git a/services/blamer/heap_objects/blame_data.h b/services/blamer/heap_objects/blame_data.h
new file mode 100644
index 0000000000000000000000000000000000000000..045bd3e4250265df81782cdfb4c87510399405ba
--- /dev/null
+++ b/services/blamer/heap_objects/blame_data.h
@@ -0,0 +1,41 @@
+// Copyright 2017 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+// Declares the hierarchy of classes that are used for the tallying of
+// performance data by the blamer service.
+
+#ifndef SERVICES_BLAMER_PUBLIC_CPP_BLAME_DATA_H_
+#ifndef SERVICES_BLAMER_PUBLIC_CPP_BLAME_DATA_H_
+
+#include <atomic>
+
+namespace blamer {
+
+// This is the data store for actual metrics gathered and attributed to a
+// particular blame node. These can be written to and read from by multiple
+// threads and processes at the same time. As such, they make use of atomics to
+// keep them thread safe. The backing memory will be in a shared memory segment
+// exported by the blame service to all other connected processes.
+struct FlatBlameData {
+ // For use with SharedMemoryHeap and PersistentMemoryAllocator.
+ constexpr HeapObjectTypes kPersistentTypeId = HeapObjectTypes::BLAME_DATA;
+
+ // Cumulative CPU time in milliseconds. This can represent 50 days of
+ // cumulative time.
+ std::atomic<uint32_t> cumulative_task_time_ms;
+
+ // The total number of tasks that have been executed under this node.
+ std::atomic<uint32_t> task_count;
+};
+
+// TODO(chrisha): Consider a distinct in-process thread-local / sequence-local
+// representation that can be written to directly without any locking or
+// atomics. This would then be "drained" to the central store that is accessible
+// from all processes. This would mean that in-process work wouldn't be visible
+// by other processes, but the moment a task completes its attribution data
+// would be available.
+
+} // namespace blamer
+
+#endif // SERVICES_BLAMER_PUBLIC_CPP_BLAME_DATA_H_

Powered by Google App Engine
This is Rietveld 408576698