|
|
Chromium Code Reviews|
Created:
4 years ago by DmitrySkiba Modified:
4 years ago CC:
chromium-reviews, tracing+reviews_chromium.org, wfh+watch_chromium.org, vmpstr+watch_chromium.org, oysteine Target Ref:
refs/pending/heads/master Project:
chromium Visibility:
Public. |
Description[tracing] Whitelist sync MDP.
Sync MDP takes longer than other whitelisted providers (~10ms on Nexus 5X),
but that shouldn't affect usability because the provider runs on dedicated
thread (SyncThread).
BUG=649065
Committed: https://crrev.com/caffce579a5c78540e15a0a280f029abb1781818
Cr-Commit-Position: refs/heads/master@{#435717}
Patch Set 1 #
Messages
Total messages: 15 (6 generated)
dskiba@chromium.org changed reviewers: + primiano@chromium.org, ssid@chromium.org
Description was changed from ========== [tracing] Whitelist sync MDP. Sync MDP takes longer than other whitelisted providers (~10ms), but that shouldn't affect usability because the provider runs on SyncThread. BUG=649065 ========== to ========== [tracing] Whitelist sync MDP. Sync MDP takes longer than other whitelisted providers (~10ms on Nexus 5X), but that shouldn't affect usability because the provider runs on dedicated thread (SyncThread). BUG=649065 ==========
On 2016/11/29 20:18:27, DmitrySkiba wrote: Any idea how come 10ms? I glanced through the code and it seems to do very little. Where does that 10ms come from? ssid: opinions on the time here? You know all the numbers.
On 2016/11/30 11:36:46, Primiano Tucci wrote: > On 2016/11/29 20:18:27, DmitrySkiba wrote: > > Any idea how come 10ms? I glanced through the code and it seems to do very > little. Where does that 10ms come from? > ssid: opinions on the time here? You know all the numbers. That's because EstimateMemoryUsage() neatly hides complexity :) In the end we end going pretty deep, for example: Directory::metahandles_map is a std::unordered_map<int64_t, std::unique_ptr<EntryKernel>>, EntryKernel::specifics_fields is a EntitySpecificsPtr[PROTO_FIELDS_COUNT], sync_pb::EntitySpecifics references ~30 other sync_pb fields, which reference others, etc.
On 2016/11/30 17:55:46, DmitrySkiba wrote: > On 2016/11/30 11:36:46, Primiano Tucci wrote: > > On 2016/11/29 20:18:27, DmitrySkiba wrote: > > > > Any idea how come 10ms? I glanced through the code and it seems to do very > > little. Where does that 10ms come from? > > ssid: opinions on the time here? You know all the numbers. Currently all the dump providers together typically takes 5-10ms in each process. This would probably double this overhead. But, we do not have any other way to get data about sync usage.
LGTM as per internal discussion due to the fact that this happens on another thread, assuming ssid is fine with this. However, feels like we can do something better / easier for sync. 10 ms of computation is quite a lot.
also please keep always +oysteine@ in the loop (read: CC him) when you make changes to things that affect slow reports.
The CQ bit was checked by dskiba@chromium.org
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...
CQ is committing da patch.
Bot data: {"patchset_id": 1, "attempt_start_ts": 1480623417268200, "parent_rev":
"f51dc3850f78a5422d16865cb60668ac5e1ee2bc", "commit_rev":
"4e84f4aeccbf49a0d33ae9d33f533f5178616a80"}
Message was sent while issue was closed.
Description was changed from ========== [tracing] Whitelist sync MDP. Sync MDP takes longer than other whitelisted providers (~10ms on Nexus 5X), but that shouldn't affect usability because the provider runs on dedicated thread (SyncThread). BUG=649065 ========== to ========== [tracing] Whitelist sync MDP. Sync MDP takes longer than other whitelisted providers (~10ms on Nexus 5X), but that shouldn't affect usability because the provider runs on dedicated thread (SyncThread). BUG=649065 ==========
Message was sent while issue was closed.
Committed patchset #1 (id:1)
Message was sent while issue was closed.
Description was changed from ========== [tracing] Whitelist sync MDP. Sync MDP takes longer than other whitelisted providers (~10ms on Nexus 5X), but that shouldn't affect usability because the provider runs on dedicated thread (SyncThread). BUG=649065 ========== to ========== [tracing] Whitelist sync MDP. Sync MDP takes longer than other whitelisted providers (~10ms on Nexus 5X), but that shouldn't affect usability because the provider runs on dedicated thread (SyncThread). BUG=649065 Committed: https://crrev.com/caffce579a5c78540e15a0a280f029abb1781818 Cr-Commit-Position: refs/heads/master@{#435717} ==========
Message was sent while issue was closed.
Patchset 1 (id:??) landed as https://crrev.com/caffce579a5c78540e15a0a280f029abb1781818 Cr-Commit-Position: refs/heads/master@{#435717} |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
