Closed Bug 415489 Opened 17 years ago Closed 17 years ago

URL bar autocomplete slow since bug 414507

Categories

(Firefox :: Address Bar, defect, P1)

defect

Tracking

()

VERIFIED FIXED
Firefox 3 beta3

People

(Reporter: ria.klaassen, Assigned: Mardak)

References

Details

(Keywords: perf, regression)

Attachments

(1 file)

Since a few days the autocomplete is extremely slow while typing. Normally when I type I see results appear instantly with every keystroke, but now I need to wait 2 or 3 seconds before the autocomplete expands. Regression window: http://bonsai.mozilla.org/cvsquery.cgi?module=PhoenixTinderbox&date=explicit&mindate=1201587360&maxdate=1201590059 Appears to be caused by bug 414507.
I'm seeing this on Linux, too.
Flags: blocking-firefox3?
OS: Windows XP → All
Hardware: PC → All
Target Milestone: --- → Firefox 3 beta3
Could also be caused by bug 414254; is also in that range.
OS: All → Windows XP
Hardware: All → PC
Target Milestone: Firefox 3 beta3 → ---
OS: Windows XP → All
Hardware: PC → All
Target Milestone: --- → Firefox 3 beta3
Sorry Reed, for wiping out your changes. The server didn't warn me.
Keywords: perf
Yeah, probably is bug 414254 if the things you're matching aren't recent. Can you try changing these prefs.. browser.urlbar.search.chunkSize to 1000 browser.urlbar.search.timeout to 50
It helps but is still a lot slower. The difference is that before 28 Jan the autocomplete stayed expanded and the results changed instantly and after 29 Jan the dropdown is collapsed as long as you are typing and only expands when you stop. Yesterday I was shopping for "schilderijverlichting" and after 30 to 100 visits to different sites I wanted to revisit certain sites and I was quite surprised that I had to wait for the dropdown after every character I typed. My places.sqlite is about 13,5 MB.
Before, it would search through your whole history before coming back with a number of results, and now it looks at a number of pages per chunk. You can make the chunkSize even larger like 5000 if you want. (I've got 14k pages total in my db.) Also, try tweaking maxRichResults browser.urlbar.search.chunkSize 5000 browser.urlbar.search.timeout 50 browser.urlbar.maxRichResults 10
Seth, Dietrich: What /is/ the timeout for anyway? Is it just so that the thread gives its priority to other threads like the UI? Do we really need it to wait anything more than say.. 10ms? I suppose one other use of the timeout is so that the user can type in characters in between the chunks, but that's not as important now that we've got bug 414254 fixed. As for chunk size, 1000 or 2000 seem good from people on IRC and here. Before we switched to this form of chunking, we would be going through /all pages/ (10k+) within a second, so a thousand should be easy. We can tweak the max rich results as well, but that kinda depends on bug 406257.
(In reply to comment #6) > Before, it would search through your whole history before coming back with a > number of results, and now it looks at a number of pages per chunk. You can > make the chunkSize even larger like 5000 if you want. (I've got 14k pages total > in my db.) > > Also, try tweaking maxRichResults > > browser.urlbar.search.chunkSize 5000 > browser.urlbar.search.timeout 50 With those two settings the speed is comparable with the speed before 28 January, in the particular case I mentioned in comment 5. Don't know exactly. Wouldn't it be safer to change the defaults and to document this in Help, or even in Options?
Regressing performance on the location bar search for a public beta is, IMO, totally not OK. Blocking for consideration. Hopefully we can solve this by tweaking the prefs. (In reply to comment #8) > With those two settings the speed is comparable with the speed before 28 > January, in the particular case I mentioned in comment 5. Don't know exactly. > Wouldn't it be safer to change the defaults and to document this in Help, or > even in Options? Well, if Ed wants to blog about this a little so that people know what settings they can tweak, that would be great. I don't think we want user-visible options on this, though.
Flags: blocking-firefox3? → blocking-firefox3+
Priority: -- → P1
1000 / 50 seems right to me. On my wee little G4 powerbook the difference between 1k and 2k seems tiny, so I'd go lower. 50 vs. 10 doesn't seem discernible. Let's get a patch in ASAP.
Attached patch v1 (deleted) — Splinter Review
2000/50 works fine on my iBook G4 1.33GHz, but I do also shrink the number of max results. We'll go with 1000/50 for now.
Assignee: nobody → edilee
Status: NEW → ASSIGNED
Attachment #301313 - Flags: review?(dietrich)
Comment on attachment 301313 [details] [diff] [review] v1 r=me for incremental improvement. for b4 we should look further into fixing comment #5.
Attachment #301313 - Flags: review?(dietrich) → review+
Attachment #301313 - Flags: approval1.9b3?
Comment on attachment 301313 [details] [diff] [review] v1 a=beltzner
Attachment #301313 - Flags: approval1.9b3? → approval1.9b3+
Checking in browser/app/profile/firefox.js; /cvsroot/mozilla/browser/app/profile/firefox.js,v <-- firefox.js new revision: 1.274; previous revision: 1.273 done
Status: ASSIGNED → RESOLVED
Closed: 17 years ago
Resolution: --- → FIXED
*** VERIFIED Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9b3) Gecko/2008020514 Firefox/3.0b3
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: