1287027 - (gdoc_read_utf8_txt_1_win7(88.18%)) [google suite][google docs] FirstPaint/FirstVisualChange much slower loading a 1 page (UTF-8) google doc

Status: RESOLVED INCOMPLETE

(Core :: Performance, defect, P1)

Description

[Walter Chen[:ypwalter][:wachen]]

# Test Case	
STR
1. Launch the browser with blank page
2. Open the google doc with 1 page (UTF-8) content (https://goo.gl/QgcVcP)
3. close the browser

# Hardware
OS: Windows 7
CPU: i7-3770 3.4GMhz
Memory: 16GB Ram
Hard Drive: 1TB SATA HDD
Graphics: GK107 [GeForce GT 640]/ GF108 [GeForce GT 440/630]

# Browsers
Firefox version: 47
Chrome version: 51.0.2704.103

# Result
Browser | Run time (median value) 
Firefox | 9811 ms
Chrome  | 5214 ms

Comment 1

[Daniel Holbert [:dholbert]]

Here's profile of loading that google doc:
http://bit.ly/2Seek7f

Comment 2

[Daniel Holbert [:dholbert]]

Here's another profile with screenshots included:
http://bit.ly/2Xhl8Ve

It shows that the doc looks visually loaded about 3 seconds after I've started loading the page (though there's work that keeps happening after that).

Comment 3

[Markus Stange [:mstange]]

This is a rather old bug, it would be great to have up-to-date numbers that compare this page between Chrome and Firefox. If I load this page manually on my machine, just eyeballing it it seems to take about the same time in both browsers before the relevant content is on the screen.
Denis, do you still have a setup that lets you easily do such comparisons?

Comment 4

[Denis Palmeiro [:denispal]]

I tested this on the 2017 Acer. Overall, it looks like we're competetive against Chrome when it comes to just loading the page. However, our first paint and first visual change numbers are very bad compared to Chrome. We should probably keep track of this bug after more of the scheduling work is landed.

Firefox Nightly:

1. INFO: https://goo.gl/QgcVcP BackEndTime: 1455 DomContentLoadedTime: 2978 FirstPaint: 2952 PageLoadTime: 7363
   INFO: VisualMetrics FirstVisualChange: 3017 SpeedIndex: 3225 LastVisualChange: 17467
2. INFO: https://goo.gl/QgcVcP BackEndTime: 1227 DomContentLoadedTime: 2769 FirstPaint: 2755 PageLoadTime: 7705
   INFO: VisualMetrics FirstVisualChange: 2817 SpeedIndex: 3048 LastVisualChange: 17317
3. INFO: https://goo.gl/QgcVcP BackEndTime: 2333 DomContentLoadedTime: 3863 FirstPaint: 3835 PageLoadTime: 8141
   INFO: VisualMetrics FirstVisualChange: 3900 SpeedIndex: 4088 LastVisualChange: 17817

Chrome:

1. INFO: https://goo.gl/QgcVcP BackEndTime: 611 DomContentLoadedTime: 4746 FirstPaint: 1839 PageLoadTime: 8681
   INFO: VisualMetrics FirstVisualChange: 2000 SpeedIndex: 3412 LastVisualChange: 18416
2. INFO: https://goo.gl/QgcVcP BackEndTime: 1210 DomContentLoadedTime: 4429 FirstPaint: 2392 PageLoadTime: 8475
   INFO: VisualMetrics FirstVisualChange: 2550 SpeedIndex: 4372 LastVisualChange: 18550
3. INFO: https://goo.gl/QgcVcP BackEndTime: 754 DomContentLoadedTime: 3018 FirstPaint: 1162 PageLoadTime: 7086
   INFO: VisualMetrics FirstVisualChange: 1350 SpeedIndex: 2958 LastVisualChange: 17133

Comment 5

[Massimo Fidanza]

(In reply to Denis Palmeiro [:denispal] from comment #4)

I tested this on the 2017 Acer. Overall, it looks like we're competetive against Chrome when it comes to just loading the page. However, our first paint and first visual change numbers are very bad compared to Chrome. We should probably keep track of this bug after more of the scheduling work is landed.
When scheduling work will be landed?

Firefox Nightly:

1. INFO: https://goo.gl/QgcVcP BackEndTime: 1455 DomContentLoadedTime: 2978 FirstPaint: 2952 PageLoadTime: 7363
   INFO: VisualMetrics FirstVisualChange: 3017 SpeedIndex: 3225 LastVisualChange: 17467
2. INFO: https://goo.gl/QgcVcP BackEndTime: 1227 DomContentLoadedTime: 2769 FirstPaint: 2755 PageLoadTime: 7705
   INFO: VisualMetrics FirstVisualChange: 2817 SpeedIndex: 3048 LastVisualChange: 17317
3. INFO: https://goo.gl/QgcVcP BackEndTime: 2333 DomContentLoadedTime: 3863 FirstPaint: 3835 PageLoadTime: 8141
   INFO: VisualMetrics FirstVisualChange: 3900 SpeedIndex: 4088 LastVisualChange: 17817

Chrome:

1. INFO: https://goo.gl/QgcVcP BackEndTime: 611 DomContentLoadedTime: 4746 FirstPaint: 1839 PageLoadTime: 8681
   INFO: VisualMetrics FirstVisualChange: 2000 SpeedIndex: 3412 LastVisualChange: 18416
2. INFO: https://goo.gl/QgcVcP BackEndTime: 1210 DomContentLoadedTime: 4429 FirstPaint: 2392 PageLoadTime: 8475
   INFO: VisualMetrics FirstVisualChange: 2550 SpeedIndex: 4372 LastVisualChange: 18550
3. INFO: https://goo.gl/QgcVcP BackEndTime: 754 DomContentLoadedTime: 3018 FirstPaint: 1162 PageLoadTime: 7086
   INFO: VisualMetrics FirstVisualChange: 1350 SpeedIndex: 2958 LastVisualChange: 17133

Comment 6

[Massimo Fidanza]

When scheduling work will be landed?

Comment 7

[Denis Palmeiro [:denispal]]

(In reply to Massimo Fidanza from comment #6)

When scheduling work will be landed?

There is no planned date as it's a work in progress at the moment.

Comment 8

[Randell Jesup [:jesup] (needinfo me)]

Moving to performance until we know why it's slow on FP/etc

Comment 9

[Mike Conley (:mconley) (:⚙️)]

Setting to P1 based on the calculator, but the first step here is getting more recent numbers and profiles to find the bottleneck.

Comment 10

[Frank Doty [:fdoty]]

Document linked in the profile no longer exists - if a new example can be produced, the bug can be reopened / a new bug can be created for tracking.