634097 - Compare beta 10 hardware acceleration usage to beta 11 hardware acceleration usage

Status: RESOLVED FIXED

(Socorro :: Data request, task)

Description

[Joe Drew (not getting mail)]

We've got some complaints about beta 11 being slower than beta 10, and from the graphics POV, this might be because we're now being much more aggressive about blocking old drivers.

To see whether this actually resulted in us having significantly fewer users of hardware acceleration, we'd like to check what proportion of our Windows Vista & 7 users are running with D2D in beta 11, and compare that to Beta 10.

Comment 1

[chris hofmann]

I'm not sure to how to get at the ratio's, but its easy to get counts of number of crash reports that have various combos of OS Ver  Adapter and Driver Info, or some buckets of those for beta 11 and b10.  

top combos of these for b11 are shown below.  counts on the left

 643 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.2202
 444 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a02, AdapterDriverVersion: 8.15.10.1930
 339 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 29c2, AdapterDriverVersion: 8.15.10.1930
 306 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 0046, AdapterDriverVersion: 8.15.10.2202
 251 6.1.7600 \N
 178 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 29c2, AdapterDriverVersion: 8.15.10.1749
 171 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 27ae, AdapterDriverVersion: 8.15.10.1930
 166 6.0.6002 Service Pack 2 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.2202
 158 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: a011, AdapterDriverVersion: 8.14.10.2117
 156 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2772, AdapterDriverVersion: 8.15.10.1930
 152 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 27a2, AdapterDriverVersion: 8.15.10.1930
...

and for beta 10 the counts look like this.  (we don't have driver info as accessble there.)


 383 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.2202 
 289 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.1749 
 280 6.1.7600 \N
 274 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 29c2, AdapterDriverVersion: 8.15.10.1930 
 267 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a02, AdapterDriverVersion: 8.15.10.1930 
 249 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 0046, AdapterDriverVersion: 8.15.10.2202 
 161 6.1.7600 AdapterVendorID: 1002, AdapterDeviceID: 68b8, AdapterDriverVersion: 8.801.0.0 
 139 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.1892 
 131 6.1.7600 AdapterVendorID: 1002, AdapterDeviceID: 9612, AdapterDriverVersion: 8.632.1.2000 
 130 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2772, AdapterDriverVersion: 8.15.10.1930 
 130 6.1.7600 AdapterVendorID: 1002, AdapterDeviceID: 95c4, AdapterDriverVersion: 8.632.1.2000 
 129 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 27ae, AdapterDriverVersion: 8.15.10.1930 
 128 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0e22, AdapterDriverVersion: 8.17.12.6658 
 125 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 27a2, AdapterDriverVersion: 8.15.10.1930 
 122 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0640, AdapterDriverVersion: 8.17.12.5896 
 119 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0640, AdapterDriverVersion: 8.17.12.6099 
 113 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0622, AdapterDriverVersion: 8.17.12.6099 
 110 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.1808 
 110 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 29c2, AdapterDriverVersion: 8.15.10.1749 
 107 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0640, AdapterDriverVersion: 8.15.11.8593 
  99 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a02, AdapterDriverVersion: 8.15.10.1749 
  97 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.1883 
  95 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: 2a42, AdapterDriverVersion: 8.15.10.2086 
  92 6.1.7600 AdapterVendorID: 8086, AdapterDeviceID: a011, AdapterDriverVersion: 8.14.10.2117 
  89 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0622, AdapterDriverVersion: 8.17.12.5896 
  87 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0e22, AdapterDriverVersion: 8.17.12.6099 
  87 6.1.7600 AdapterVendorID: 1002, AdapterDeviceID: 9442, AdapterDriverVersion: 8.801.0.0 
  81 6.1.7600 AdapterVendorID: 10de, AdapterDeviceID: 0641, AdapterDriverVersion: 8.17.12.5896 
  80 6.0.6002 Service Pack 2 AdapterVendorID: 8086, AdapterDeviceID: 2a02, AdapterDriverVersion: 7.14.10.1437 

I guess we could also look at the list of drivers we are blocking and see how they related to crash volume in b10 and b11.

Comment 2

[chris hofmann]

counts above are shown for b11 on feb 13, and b10 on jan 31 when each of the relases had about the same number active daily users (between 1.4 and 1.6 million)

Comment 3

[Bas Schouten (:bas.schouten)]

The important figure here is how many users have the d2d1.dll module loaded into their address space.

This is the only way we know if Direct2D is actually enabled for a user.

Comment 4

[chris hofmann]

ok, I've got an idea to try.

I'll get a sample of 1000 reports for 4.0b10 on jan31 and look for d2d1.dll in the module list, then do the same for 4.0b11 on feb 13.

then we can use that as rough estimates of d2d enabled v. not present at time of crash.

Comment 5

[chris hofmann]

The full 1000 report sample hasn't finished up but in a preliminary look it appears that about 25% of all crash reports on 4.0b10 had d2d1.dll in the process list, and about 8% of 4.0b11 show d2d1.dll running.

In this first run just taken a sample across all os versions, but it sounds like I should tighten up the sample to just include:

  Firefox 4.0b10 && windows_vista (6.0) && windows_7 (6.1)

  Firefox 4.0b11 && windows_vista (6.0) && windows_7 (6.1)

Comment 6

[chris hofmann]

ok, the first run of 1000 reports with the only restriction being firefox version has completed. with 88 reports containing d2d1.dll on 4.0b11 and 255 reports in the sample from 4.0b10

 d2d1-reports-for-40b11.txt  1000	88	0.088
 d2d1-reports-for-40b10.txt  1000	255	0.255

Comment 7

[chris hofmann]

If restrict the sample of 1000 reports to only include windows versions 6.[01] the numbers look like this:

d2d1-win6-reports-for-40b10  1000	606	0.606
d2d1-win6-reports-for-40b11  1000	266	0.266

Comment 8

[Laura Thomson :laura]

chofmann: do you have all the data you need?  Do you need anything from PG/HBase?

Comment 9

[chris hofmann]

I took samples of b8  and b9 a few days after there release when they had about 1.4 million active daily users.  Here is what those numbers look like:

                                            ratio where d2d1.dll found

d2d1-win6-reports-for-40b8  jan12  1000	433    0.433
d2d1-win6-reports-for-40b9  jan21  1000	456    0.456
d2d1-win6-reports-for-40b10 jan31  1000 606    0.606
d2d1-win6-reports-for-40b11 feb13  1000 266    0.266

Comment 10

[Robert Kaiser]

Is this data request still relevant, i.e. does it still need to be done? If not, can you close the bug? Thanks.

Comment 11

[Will Kahn-Greene [:willkg] ET needinfo? me]

I'm closing this out per comment #10.