1150768 - WebGL conformance 1.0.2 / 1.0.3: regression on textures/copy-tex-image-2d-formats.html fails on Intel HD 3000 on Windows 7 / 10

Status: NEW

(Core :: Graphics: CanvasWebGL, defect, P3)

Description

[Luke]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:40.0) Gecko/20100101 Firefox/40.0
Build ID: 20150402115824

Steps to reproduce:

With Intel GPU
Steps to reproduce:

1. With an Intel HD Graphics GPU visit: https://www.khronos.org/registry/webgl/conformance-suites/1.0.2/conformance/textures/copy-tex-image-2d-formats.html


Actual results:

    failed: getError expected: NO_ERROR. Was OUT_OF_MEMORY : should be able to copyTexImage2D LUMINANCE_ALPHA from RGBA
    failed: should be 64,64,64,127
    failed: getError expected: NO_ERROR. Was OUT_OF_MEMORY : should be able to copyTexImage2D LUMINANCE_ALPHA from RGBA
    failed: should be 64,64,64,127


Expected results:

In older version of FF this test used to pass. See Bug 896693. The regression was fixed on Nvidia hardware, but Intel HD 4000 is still failing. Google Chrome passes this test on identical hardware. With FF both the D3D9 and D3D11 backends fail.

Graphics
Adapter Description	Intel(R) HD Graphics
Adapter Drivers	igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
Adapter RAM	Unknown
Asynchronous Pan/Zoom	none
Device ID	0x0f31
Direct2D Enabled	true
DirectWrite Enabled	true (6.1.7600.16385)
Driver Date	9-30-2014
Driver Version	10.18.10.3958
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	0f311849
Vendor ID	0x8086
WebGL Renderer	Google Inc. -- ANGLE (Intel(R) HD Graphics Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d
AzureContentBackend	direct2d
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

Comment 1

[Jukka Jylänki]

Reproduced this test to fail on the following hardware:

MACBOOK_AIR_WIN
---------------

MacBook Air (13-inch, Mid 2011)
Windows 7 Home Premium 64-bit in dual-boot mode (bootcamp)
1.7GHz Intel Core i5
4GB 1333 MHz DDR3
1440x900 pixels display
System DirectX version: DirectX 11.0
Intel HD Graphics 3000 384MB:
	- Vendor: Intel (0x8086)
	- Device ID: 0x0116
	- Revision ID: 0x0009
	- igdumd64.dll, igd10umd64.dll:
		- version 9.17.10.3517
		- date 3/20/2014
		- DDI Version 10.1, WDDM 1.1, WHQL approved
Firefox Nightly 41.0a1 (2015-06-29)

Comment 2

[Jukka Jylänki]

The failure did not occur on any of the following: SPARK, GIADA, MACBOOK_AIR_OSX, MACBOOK_PRO_OSX, MACBOOK_PRO_WIN, MACMINI, MACPRO, NEXUS-4, NEXUS-5, NEXUS-10, SURFACE, WINDBOX, HASWELL, HPOMEN (see https://bugzilla.mozilla.org/show_bug.cgi?id=1178601 for hardware configuration details of these systems)

Comment 3

[Luke]

With the d3d9 backend, this test is still failing under Windows 10 on nightly, 47.0a1 (2016-02-04). The d3d11 backend passes.

Comment 4

[Luke]

Also affects WebGL 1.03 and newer Intel hardware such as HD 5000. Confirmed on 52.0a1 (2016-10-07).

Comment 5

[Milan Sreckovic [:milan] (needinfo for best results)]

(In reply to Luke from comment #3)
> With the d3d9 backend, this test is still failing under Windows 10 on
> nightly, 47.0a1 (2016-02-04). The d3d11 backend passes.

What do you mean by D3D9 backend?  WebGL renderer reporting ANGLE D3D9 or something else?

This passes for me on Windows 10 with:
Description	Intel(R) HD Graphics 530
Vendor ID	0x8086
Device ID	0x191b
Driver Version	20.19.15.4454
Driver Date	5-4-2016

so it seems to be not all Intel that fail.

Jukka, do you still have the system from comment 1, and does the problem exist in the latest nightly?

Luke, do you still see the problem in the latest 52?  A lot has changed since October.

Comment 6

[Luke]

As I already said this is working on newer Intel GPUs with the D3D11 backend. There is no need to repeat this information as it just adds noise to the bug tracker.

It is only failing on older hardware that uses that D3D9 backend. This can be simulated with the webgl.angle.try-d3d11:false setting.

Still FAILING on 54.0a1 (2017-02-09) (64-bit)

Description	Intel(R) HD Graphics 5500
Vendor ID	0x8086
Device ID	0x1616
Driver Version	20.19.15.4549
Driver Date	11-10-2016

Comment 7

[Jukka Jylänki]

Tested this today on FF Nightly 55.0a1 (2017-03-20) (64-bit) on MACBOOK_AIR_WIN which is still running Windows 7 Home Premium, the test does pass and about:support says it is using D3D11 backend.

Do we still support the D3D9 backend after Firefox 52 when Windows XP users became legacy? Or is it now D3D11 all the way? Do we have metrics of how many users are not on D3D11 backend on Windows?

Comment 8

[Milan Sreckovic [:milan] (needinfo for best results)]

D3D11 after 52; we also never enabled D3D9+e10s combination.  1.6% of the users are on D3D9, ~71% on D3D11.  We blocklist about 15% of the user from getting D3D11.