1508765 - Firefox is very slow to render poster PDF

Status: NEW

(Firefox :: PDF Viewer, defect, P3)

Description

[Clemens Eisserer]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:65.0) Gecko/20100101 Firefox/65.0

Steps to reproduce:

1. opened PDF  located at: https://wiki.eclipse.org/images/0/07/CDO-Poster.pdf


Actual results:

Firefox / PDF.js takes 5minutes on rather modern hardware (I5-6300U / Windows-7) to render the PDF


Expected results:

Firefox should be at least as fast as evince (15s) loading this file.

Google Chrome loads the PDF in 2-3s.

Comment 1

[Clemens Eisserer]

Attachment: CDO-Poster.7z

Comment 2

[Clemens Eisserer]

uploading the gecko profile failed multiple times (XHR onerror was called), please find the gecko profiler data at: https://drive.google.com/open?id=1AmP5_Qh8vXEcbuBrcvzqDWvHzHg9dTlr

it seems a lot of time is spent inside canvas functions.

Comment 3

[Calixte Denizet (:calixte)]

Around 20s for me in nightly 85 on Linux.
:Clemens, is it still an issue for you ?

Comment 4

[Clemens Eisserer]

I guess the question is not so much whether this specific case is an issue for me, but rather why Firefox's integrated PDF viewer is about 10x slower than its direct competitor Google Chrome.

Comment 5

[Calixte Denizet (:calixte)]

Here's a profile from nightly on Windows 11:
https://share.firefox.dev/3OKBJy7

We spend a lot of time in removing an Observer:
https://searchfox.org/mozilla-central/rev/f4d3fe187cf7dffa4c13b354bbde9bc47b5ccd3f/dom/canvas/ImageBitmap.cpp#86

:tnikkel, could you have a look please ?

Comment 6

[Florian Quèze [:florian]]

It looks like this observer was added in bug 1414825.

Comment 7

[Timothy Nikkel (:tnikkel)]

It looks like nsContentUtils::RegisterShutdownObserver, and in general things that add an observer to the observer service, just store the observer in a regular array, so removing an observer is O(n) in number of observers. That is not suitable to use for objects like ImageBitmap where we might create a lot of them.

Comment 8

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1414825

Comment 9

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1414825

Comment 10

[Timothy Nikkel (:tnikkel)]

I'm gonna move that part of this that is a regression from bug 1414825 to bug 1839286, because after fixing that this testcase is still pretty slow, so there is mor going on here than just that.

Comment 11

[Andrej (:aglavic)]

The Performance Impact Calculator has determined this bug's performance impact to be medium. If you'd like to request re-triage, you can reset the Performance Impact flag to "?" or needinfo the triage sheriff.

Platforms: [x] Windows [x] macOS
Impact on site: Causes noticeable jank
Page load impact: Severe
Websites affected: Rare
[x] Affects animation smoothness
[x] Able to reproduce locally