nagai/fune
3
0
Fork 0
forked from mirrors/gecko-dev
Code Pull requests Activity

Bug 1792500 - Document inaccuracy in MEMORY_TOTAL telemetry r=kmag

Browse source 
Differential Revision: https://phabricator.services.mozilla.com/D158174
This commit is contained in:
Paul Bone 2022-09-27 06:14:57 +00:00
parent 6513faf43f
commit 6bd2b5eae8
2 changed files with 7 additions and 1 deletions
repo.diff.show_diff_stats Download patch file Download diff file Expand all files Collapse all files

2
toolkit/components/telemetry/Histograms.json
View file

@ -1234,7 +1234,7 @@
"low": 32768, "low": 32768,
"high": 16777216, "high": 16777216,
"n_buckets": 100, "n_buckets": 100,
"description": "Total Memory Across All Processes (KB)", "description": "Total Memory Across All Processes (KB) (inaccurate WRT shared memory. See MemoryTelemetry.cpp)",
"releaseChannelCollection": "opt-out" "releaseChannelCollection": "opt-out"
}, },
"MEMORY_DISTRIBUTION_AMONG_CONTENT": { "MEMORY_DISTRIBUTION_AMONG_CONTENT": {

6
xpcom/base/MemoryTelemetry.cpp
View file

@ -404,6 +404,12 @@ nsresult MemoryTelemetry::FinishGatheringTotalMemory(
int64_t aTotalMemory, const nsTArray<int64_t>& aChildSizes) { int64_t aTotalMemory, const nsTArray<int64_t>& aChildSizes) {
mGatheringTotalMemory = false; mGatheringTotalMemory = false;
// Total memory usage can be difficult to measure both accurately and fast
// enough for telemetry (iterating memory maps can jank whole processes on
// MacOS). Therefore this shouldn't be relied on as an absolute measurement
// especially on MacOS where it double-counts shared memory. For a more
// detailed explaination see:
// https://groups.google.com/a/mozilla.org/g/dev-platform/c/WGNOtjHdsdA
HandleMemoryReport(Telemetry::MEMORY_TOTAL, nsIMemoryReporter::UNITS_BYTES, HandleMemoryReport(Telemetry::MEMORY_TOTAL, nsIMemoryReporter::UNITS_BYTES,
aTotalMemory); aTotalMemory);