1256960 - Input latency when drawing on WebGL with D3D11

Status: NEW

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

Description

[pinggi]

Attachment: lag.avi

User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:45.0) Gecko/20100101 Firefox/45.0
Build ID: 20160303134406

Steps to reproduce:

Try to draw a line on http://thj.no/personal/latency_test_case.html


Actual results:

The drawn line has significant delay after mouse move. Very bad latency :-(


Expected results:

The line should be smoothly drawn without a lag. The line should be immediately drawn right next to a mouse cursor. The attached video shows in the first half Chromium and in the second half Firefox. The video doesn't have high fps so it's chopped little bit, but you can see the difference. In real Chromium is really smooth without any lag while Firefox has horrible delay :-((

Firefox 45 (32 bit and 64 bit too)
Chromium 49.0.2623.87 (64-bit)
Windows 7

Comment 1

[Loic]

Attachment: Converted_file_439c87f8.mp4

Comment 2

[Loic]

regression range:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=292fbdb78dde3dee31472afc83b865e4d77891d1&tochange=79637e9bcdaa0657cd79732b095936b377e262b4

Jeff Muizelaar — Bug 1183341. Reenable ANGLE w/ D3D11 on newer Intel GPUs. r=bas

My about:support:
Adapter Description	Intel(R) HD Graphics 3000
Adapter Description (GPU #2)	NVIDIA GeForce GT 550M
Adapter Drivers	igdumd64 igd10umd64 igd10umd64 igdumd32 igd10umd32 igd10umd32
Adapter Drivers (GPU #2)	nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter RAM	Unknown
Adapter RAM (GPU #2)	2048
Asynchronous Pan/Zoom	wheel input enabled; touch input enabled
ClearType Parameters	Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 50
Device ID	0x0116
Device ID (GPU #2)	0x0df6
Direct2D Enabled	true
DirectWrite Enabled	true (6.2.9200.17568)
Driver Date	5-26-2015
Driver Date (GPU #2)	2-23-2016
Driver Version	9.17.10.4229
Driver Version (GPU #2)	10.18.13.6200
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	17121043
Subsys ID (GPU #2)	17121043
Supports Hardware H264 Decoding	Yes
Vendor ID	0x8086
Vendor ID (GPU #2)	0x10de
WebGL Renderer	Google Inc. -- ANGLE (Intel(R) HD Graphics 3000 Direct3D11 vs_4_1 ps_4_1)
windowLayerManagerRemote	true
AzureCanvasAccelerated	0
AzureCanvasBackend	direct2d 1.1
AzureContentBackend	direct2d 1.1
AzureFallbackCanvasBackend	cairo

Comment 3

[Loic]

pinggi, can you copy here the section "graphics" from the page about:support, please.

Comment 4

[Jeff Muizelaar [:jrmuizel]]

If you disable angle d3d11 does the situation improve?

Comment 5

[Loic]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #4)
> If you disable angle d3d11 does the situation improve?

Yes, it's smooth with webgl.angle.try-d3d11=false in 48.

Comment 6

[pinggi]

@Loic sure, here is my about:support:

Adapter Description	ATI Radeon HD 4290
Adapter Drivers	aticfx64 aticfx64 aticfx32 aticfx32 atiumd64 atidxx64 atiumdag atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM	624
Asynchronous Pan/Zoom	none
Device ID	0x9714
Direct2D Enabled	true
DirectWrite Enabled	true (6.2.9200.16492)
Driver Date	4-29-2013
Driver Version	8.970.100.1100
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	00000000
Supports Hardware H264 Decoding	No; DXVA2D3D9 crashes detected in the past
Vendor ID	0x1002
WebGL Renderer	Google Inc. -- ANGLE (ATI Radeon HD 4290 Direct3D11 vs_4_1 ps_4_1)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d 1.1
AzureContentBackend	direct2d 1.1
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0
(#0) Error	DXVA2D3D9 video decoding is disabled due to a previous crash.


I tried changing the option webgl.angle.try-d3d11 to false and it's big difference - smooth as chromium, maybe better!

Comment 7

[Jeff Muizelaar [:jrmuizel]]

I can reproduce this as well. I have no guesses yet as to what the problem might be.

Comment 8

[pinggi]

The problem is that when "webgl.angle.try-d3d11" is set to false, then webgl does not work with:

http://alteredqualia.com/xg/examples/forward_performance_shadowmap.html

Comment 9

[Peter Chang[:pchang]]

Need info daosheng to check

Comment 10

[Daosheng Mu[:daoshengmu]]

Here is my support, and I can't reproduce this problem. The result on my device is smooth.

Name 	Firefox
Version 	46.0.1

Graphics
Adapter Description	NVIDIA GeForce GT 730
Adapter Drivers	nvd3dumx,nvwgf2umx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um,nvwgf2um
Adapter RAM	2048
Asynchronous Pan/Zoom	none
Device ID	0x1287
Direct2D Enabled	true
DirectWrite Enabled	true (10.0.10586.0)
Driver Date	5-9-2016
Driver Version	10.18.13.6519
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	00000000
Supports Hardware H264 Decoding	Yes
Vendor ID	0x10de
WebGL Renderer	Google Inc. -- ANGLE (NVIDIA GeForce GT 730 Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d 1.1
AzureContentBackend	direct2d 1.1
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

Not sure if it would not happen on NV cards?

Comment 11

[Daosheng Mu[:daoshengmu]]

Pinggi, can you help check your Chrome Angle setting at this site?
http://www.browserleaks.com/webgl

I would like to know if your device uses d3d11 as well?

In my case, it is 
Miscellaneous :
Antialiasing	Available
ANGLE	Yes, Direct3D 11
Major Performance Caveat	No

Comment 12

[pinggi]

Yes, anytime.

It's the same for me:

Antialiasing	Available
ANGLE	Yes, Direct3D 11
Major Performance Caveat	No

if you're interested in full info:

WebGL Support Detection
Does Your Browser Supports WebGL?	✔ Yes
Is WebGL Enabled in Your Browser?	✔ Yes
What about WebGL 2.0 Support?	× No
Browser User-Agent:	Mozilla/5.0 (Windows NT 6.1; WOW64; rv:46.0) Gecko/20100101 Firefox/46.0

WebGL Context Info
Supported Context Name(s)	webgl, experimental-webgl
GL Version	WebGL 1.0
Shading Language Version	WebGL GLSL ES 1.0
Vendor	Mozilla
Renderer	Mozilla

Debug Renderer Info
Unmasked Vendor	Not Available
Unmasked Renderer	Not Available

Vertex Shader
Max Vertex Attributes	16
Max Vertex Uniform Vectors	1024
Max Vertex Texture Image Units	16
Max Varying Vectors	30
Best Float Precision	[-2127, 2127] (23)

Rasterizer
Aliased Line Width Range	[1, 1]
Aliased Point Size Range	[1, 1024]
Fragment Shader
Max Fragment Uniform Vectors	1024
Max Texture Image Units	16
Float/Int Precision:	highp/highp
Best Float Precision	[-2127, 2127] (23)

Framebuffer
Max Color Buffers:	8
RGBA Bits	[8, 8, 8, 8]
Depth / Stencil Bits:	[24, 8]
Max Render Buffer Size	8192
Max Viewport Dimensions	[16383, 16383]

Textures
Max Texture Size	8192
Max Cube Map Texture Size	8192
Max Combined Texture Image Units	32
Max Anisotropy	16

Miscellaneous :
Antialiasing	Available
ANGLE	Yes, Direct3D 11
Major Performance Caveat	No
Supported WebGL Extensions	ANGLE_instanced_arrays
	EXT_blend_minmax
	EXT_color_buffer_half_float
	EXT_frag_depth
	EXT_shader_texture_lod
	EXT_texture_filter_anisotropic
	OES_element_index_uint
	OES_standard_derivatives
	OES_texture_float
	OES_texture_float_linear
	OES_texture_half_float
	OES_texture_half_float_linear
	OES_vertex_array_object
	WEBGL_color_buffer_float
	WEBGL_compressed_texture_etc1
	WEBGL_compressed_texture_s3tc
	WEBGL_depth_texture
	WEBGL_draw_buffers
	WEBGL_lose_context
	MOZ_WEBGL_lose_context
	MOZ_WEBGL_compressed_texture_s3tc
	MOZ_WEBGL_depth_texture

Do you have any idea already what could be the cause?

Comment 13

[Daosheng Mu[:daoshengmu]]

(In reply to pinggi from comment #12)
> Yes, anytime.
> 
> It's the same for me:
> 
> Antialiasing	Available
> ANGLE	Yes, Direct3D 11
> Major Performance Caveat	No
> 
> if you're interested in full info:
> 
> WebGL Support Detection
> Does Your Browser Supports WebGL?	✔ Yes
> Is WebGL Enabled in Your Browser?	✔ Yes
> What about WebGL 2.0 Support?	× No
> Browser User-Agent:	Mozilla/5.0 (Windows NT 6.1; WOW64; rv:46.0)
> Gecko/20100101 Firefox/46.0
> 
> WebGL Context Info
> Supported Context Name(s)	webgl, experimental-webgl
> GL Version	WebGL 1.0
> Shading Language Version	WebGL GLSL ES 1.0
> Vendor	Mozilla
> Renderer	Mozilla
> 
> Debug Renderer Info
> Unmasked Vendor	Not Available
> Unmasked Renderer	Not Available
> 
> Vertex Shader
> Max Vertex Attributes	16
> Max Vertex Uniform Vectors	1024
> Max Vertex Texture Image Units	16
> Max Varying Vectors	30
> Best Float Precision	[-2127, 2127] (23)
> 
> Rasterizer
> Aliased Line Width Range	[1, 1]
> Aliased Point Size Range	[1, 1024]
> Fragment Shader
> Max Fragment Uniform Vectors	1024
> Max Texture Image Units	16
> Float/Int Precision:	highp/highp
> Best Float Precision	[-2127, 2127] (23)
> 
> Framebuffer
> Max Color Buffers:	8
> RGBA Bits	[8, 8, 8, 8]
> Depth / Stencil Bits:	[24, 8]
> Max Render Buffer Size	8192
> Max Viewport Dimensions	[16383, 16383]
> 
> Textures
> Max Texture Size	8192
> Max Cube Map Texture Size	8192
> Max Combined Texture Image Units	32
> Max Anisotropy	16
> 
> Miscellaneous :
> Antialiasing	Available
> ANGLE	Yes, Direct3D 11
> Major Performance Caveat	No
> Supported WebGL Extensions	ANGLE_instanced_arrays
> 	EXT_blend_minmax
> 	EXT_color_buffer_half_float
> 	EXT_frag_depth
> 	EXT_shader_texture_lod
> 	EXT_texture_filter_anisotropic
> 	OES_element_index_uint
> 	OES_standard_derivatives
> 	OES_texture_float
> 	OES_texture_float_linear
> 	OES_texture_half_float
> 	OES_texture_half_float_linear
> 	OES_vertex_array_object
> 	WEBGL_color_buffer_float
> 	WEBGL_compressed_texture_etc1
> 	WEBGL_compressed_texture_s3tc
> 	WEBGL_depth_texture
> 	WEBGL_draw_buffers
> 	WEBGL_lose_context
> 	MOZ_WEBGL_lose_context
> 	MOZ_WEBGL_compressed_texture_s3tc
> 	MOZ_WEBGL_depth_texture
> 
> Do you have any idea already what could be the cause?

Sorry, pinggi. Could you help me use Google Chrome to dump the info from the same link, http://www.browserleaks.com/webgl.

Thanks.

I wounder if Google Chrome doesn't enable DX11 for intel GPUs.

Comment 14

[pinggi]

Here's the Chromium output:

Antialiasing	Available
ANGLE	Yes, Direct3D 9
Major Performance Caveat	No


and the full dump:

WebGL Support Detection
Does Your Browser Supports WebGL?	✔ Yes
Is WebGL Enabled in Your Browser?	✔ Yes
What about WebGL 2.0 Support?	× No
Browser User-Agent:	Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/49.0.2623.87 Safari/537.36

WebGL Context Info
Supported Context Name(s)	webgl, experimental-webgl
GL Version	WebGL 1.0 (OpenGL ES 2.0 Chromium)
Shading Language Version	WebGL GLSL ES 1.0 (OpenGL ES GLSL ES 1.0 Chromium)
Vendor	WebKit
Renderer	WebKit WebGL

Debug Renderer Info
Unmasked Vendor	! Google Inc.
Unmasked Renderer	! ANGLE (ATI Radeon HD 4290 Direct3D9Ex vs_3_0 ps_3_0)

Vertex Shader
Max Vertex Attributes	16
Max Vertex Uniform Vectors	253
Max Vertex Texture Image Units	4
Max Varying Vectors	9
Best Float Precision	[-2127, 2127] (23)

Rasterizer
Aliased Line Width Range	[1, 1]
Aliased Point Size Range	[1, 256]

Fragment Shader
Max Fragment Uniform Vectors	221
Max Texture Image Units	16
Float/Int Precision:	highp/highp
Best Float Precision	[-2127, 2127] (23)

Framebuffer
Max Color Buffers:	1
RGBA Bits	[8, 8, 8, 8]
Depth / Stencil Bits:	[24, 8]
Max Render Buffer Size	8192
Max Viewport Dimensions	[8192, 8192]

Textures
Max Texture Size	8192
Max Cube Map Texture Size	8192
Max Combined Texture Image Units	20
Max Anisotropy	16

Miscellaneous :
Antialiasing	Available
ANGLE	Yes, Direct3D 9
Major Performance Caveat	No
Supported WebGL Extensions	ANGLE_instanced_arrays
EXT_blend_minmax
EXT_frag_depth
EXT_shader_texture_lod
EXT_texture_filter_anisotropic
WEBKIT_EXT_texture_filter_anisotropic
OES_element_index_uint
OES_standard_derivatives
OES_texture_float
OES_texture_float_linear
OES_texture_half_float
OES_texture_half_float_linear
OES_vertex_array_object
WEBGL_compressed_texture_s3tc
WEBKIT_WEBGL_compressed_texture_s3tc
WEBGL_debug_renderer_info
WEBGL_debug_shaders
WEBGL_lose_context
WEBKIT_WEBGL_lose_context

Comment 15

[pinggi]

So the difference is:

					         Chrome vs Firefox
Max Vertex Uniform Vectors		            253 vs 1024
Max Vertex Texture Image Units		              4 vs 16
Max Varying Vectors			              9 vs 30
Aliased Point Size Range		       [1, 256] vs [1, 1024]
Max Fragment Uniform Vectors		            221 vs 1024
Max Color Buffers:			              1 vs 8
Max Viewport Dimensions			   [8192, 8192] vs [16383, 16383]
Max Combined Texture Image Units	             20 vs 32
ANGLE					Yes, Direct3D 9 vs Yes, Direct3D 11

Supported WebGL Extensions

Unique for Chrome

	WEBKIT_EXT_texture_filter_anisotropic
	WEBKIT_WEBGL_compressed_texture_s3tc
	WEBGL_debug_renderer_info
	WEBGL_debug_shaders
	WEBKIT_WEBGL_lose_context

Unique for Firefox

	EXT_color_buffer_half_float
	WEBGL_compressed_texture_etc1
	MOZ_WEBGL_compressed_texture_s3tc
	WEBGL_depth_texture
	WEBGL_draw_buffers
	MOZ_WEBGL_lose_context
	MOZ_WEBGL_depth_texture

Comment 16

[pinggi]

It seems that Firefox adds more overhead by using Direct3D 11 or is there something else odd?
Is it possible to set WebGL in Firefox to use Direct3D 9?

Comment 17

[pinggi]

Mmm, I didi second attempt to set "webgl.angle.try-d3d11" to "false", restarted Firefox and now Firefox uses Direct3D 9. The only difference on http://www.browserleaks.com/webgl between Firefox and Chrome is in Supported WebGL Extensions:

Firefox:
	EXT_color_buffer_half_float
	WEBGL_color_buffer_float
	MOZ_WEBGL_lose_context
	MOZ_WEBGL_compressed_texture_s3tc

Chrome:
	WEBKIT_EXT_texture_filter_anisotropic
	WEBKIT_WEBGL_compressed_texture_s3tc
	WEBGL_debug_renderer_info
	WEBGL_debug_shaders
	WEBKIT_WEBGL_lose_context

I tested both sites...

	http://thj.no/personal/latency_test_case.html
	http://alteredqualia.com/xg/examples/forward_performance_shadowmap.html

...again with surprise that all is working now. The first one is smooth and the second one work. So now I'm using WebGL with Direct3D 9.

So definitely WebGL with Direct3D 11 introduces some latency on the latency test web page.

Comment 18

[Jeff Muizelaar [:jrmuizel]]

It's probably worth trying to disable shared surface support and see if the problem still happens.
Setting keyedMutex to null here: https://dxr.mozilla.org/mozilla-central/rev/c4449eab07d39e20ea315603f1b1863eeed7dcfe/gfx/gl/SharedSurfaceANGLE.cpp#73 should do the trick.

Comment 19

[Daosheng Mu[:daoshengmu]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #18)
> It's probably worth trying to disable shared surface support and see if the
> problem still happens.
> Setting keyedMutex to null here:
> https://dxr.mozilla.org/mozilla-central/rev/
> c4449eab07d39e20ea315603f1b1863eeed7dcfe/gfx/gl/SharedSurfaceANGLE.cpp#73
> should do the trick.

I have tried to set keyedMutex to null on my NVIDIA-PC that uses Angle-Dx11. My canvas is gray that seems to ANGLE sharedsurface does not be ready to copy to my framebuffer.

Comment 20

[Peter Chang[:pchang]]

Change tile based on comment 17.

Comment 21

[Peter Chang[:pchang]]

I can't reproduce this issue with Intel HD 4000. Jeff, which video card you use to reproduce this issue? I saw pinggi is using ATI 4290.

Comment 22

[Daosheng Mu[:daoshengmu]]

I can't reproduce with Windows Intel HD 4600 as well. By the way, I check this in Chromium. It uses ANGLE (Intel(R) HD Graphics 4600 Direct3D11 vs_5_0 ps_5_0) that is the same to Firefox 46. Therefore, I don't think it would not be a Dx11 issue. It is ATI and HD 3000 Dx11 driver issues.

Comment 23

[Jeff Muizelaar [:jrmuizel]]

(In reply to Peter Chang[:pchang] from comment #21)
> I can't reproduce this issue with Intel HD 4000. Jeff, which video card you
> use to reproduce this issue? I saw pinggi is using ATI 4290.

I can reproduce this on a Intel HD 4000 on Windows 10 with the 10.18.10.4276 driver. Can you reproduce in that configuration?

Comment 24

[Daosheng Mu[:daoshengmu]]

I rollback my driver of Intel HD 4600 on Windows 10 to 10.18.10.4276. It still can't be reproduced...Btw, it uses Dx11 as well.

Comment 25

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

Jeff, can we still reproduce this?

Comment 26

[Liz Henry (:lizzard) (relman/hg->git project)]

Andrei can your team reproduce this with the older driver?
Milan, might we want to block the older driver if so?

Comment 27

[Alexandru Simonca [:asimonca], Release Desktop QA]

Hi Liz, 
I have reproduced this issue using Firefox RC 45.0 build 1 (id: 20160301003640) on a Windows 7 Pro (x64) machine that uses ATI RADEON 3000 graphics card and has the 8.970.100.3000 driver and Direct3D 11.
Please ni? me if further testing is needed on the issue.

Comment 28

[Kartikaya Gupta (email:[e-mail])]

I talked to Jeff about this, right now it's still not clear why it reproduces in some places and not others. In order to figure this out, we need more information on what configurations it reproduces on and what configurations it doesn't reproduce on.

Andrei, could you check on a range of hardware to see what configurations this issue can be reproduced on, and which ones don't have the problem? Hopefully that will give us some idea of what the pattern is.

Comment 29

[Alexandru Simonca [:asimonca], Release Desktop QA]

Hello,
I have tested some more hardware for this issue and it seems that the problem is on ATI graphics cards that have the driver version under "10". Since we are limited on the hardware that we have here I've managed to get the following results:

1. Win 7 x64 - GPU AMD 760G - driver 8.723.0.0 - Firefox 45.0 RC & Nightly 51 - REPRODUCIBLE
2. Win XP x64 - GPU AMD R7 - driver 13.251.0.0 - Firefox 45.0 RC & Nightly 51 - Not reproducible

Comment 30

[Kartikaya Gupta (email:[e-mail])]

Jeff, does comment 29 help any?

Comment 31

[Ryan VanderMeulen [:RyanVM]]

Milan, can you help move this along?

Comment 32

[Liz Henry (:lizzard) (relman/hg->git project)]

Too late for a fix for 53, as we are in the last week of the 53 beta cycle.

Comment 33

[Gerry Chang [:gchang]]

Too late for 54, let's see if we can fix it in 55.

Comment 34

[Kartikaya Gupta (email:[e-mail])]

Dropping stale needinfo