Cool VL Viewer forum

View unanswered posts | View active topics It is currently 2024-03-19 07:56:03



Reply to topic  [ 29 posts ]  Go to page 1, 2, 3  Next
Experimental viewer v1.29.0 feedback thread 
Author Message

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
This thread is to be used for feedback about the v1.29.0 experimental viewer branch.

Let's start with today's release (v1.29.0.2) new threaded GL image worker.

Here are the known glitches (potentially) seen with this feature and their workarounds:

  • The feature does not properly work with Intel iGPUs (taken into account by LL for potential future resolution). EDIT: this has proven untrue for me. Feel free to report any failure you would experience, quoting your hardware (iGPU model) and software (OS, drivers).
  • You could potentially (*) see flashing (black) and/or transparent/bogus textures; should this happen, setting the new "GLWorkerUseFence" debug setting to TRUE should solve the problem (at the cost of slower GL image creations). EDIT: workaround removed from v1.29.0.6 since no one has reported needing it.
  • You could potentially (*) see issues with bump map textures rendering, especially in ALM mode. LL disabled bump maps threaded creation in their "performance viewer", but they probably saw glitches because of one of the bugs I fixed in their code. Changing the new "GLWorkerUseForBumpMaps" debug setting to FALSE should solve the problem when seen happening. EDIT: workaround removed from v1.29.0.6 since no one has reported needing it.
  • Streaming media textures using gstreamer had to be made non-threaded, due to flashing happening on media textures after a couple minutes of media playback (unlike VLC, used by LL, gstreamer is multi-threaded which seems to produce weird synchronization glitches). There is also a debug setting to control this ("GLWorkerUseForStreamingMedia"), but it should stay FALSE; not a big deal since there is no real performances impact.
  • Under Linux (not Windows), with NVIDIA's official drivers, I am seeing "blocky" login screen UI elements/fonts for two or three seconds at viewer launch; I could not figure out why this happens even though I only use GL threading after login (so, the viewer should render exactly the same before login as when GL threading is disabled). I am interested to know whether you are seeing it or not with other hardware/drivers combinations (please report it here and quote your GPU/driver configuration). This is a cosmetic and harmless glitch, but I'd like to fix it... EDIT: fixed.


(*) Normally, you should not see this issue happening at all thanks to the bug fixes I brought to LL's code. If you see this issue happening, please do report it here and quote your hardware/software (GPU/OS/driver) configuration so that I can investigate further.


2022-04-16 09:31:01
Profile WWW

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
One more thing I just noticed: the threaded GL worker does not work when changing video modes because LL's code does not redo the shared contexts while the window they are associated with is destroyed during the video mode change !...

I will have to implement the proper code for this. EDIT: code ready for next release (v1.29.0.3). Problem solved ! 8-)

In the mean time, do not change the video modes settings (this includes anti-aliasing and anisotropic filtering, which trigger a GL mode change) while the viewer is logged in, and do not use either the "full screen" (non-windowed) mode, since it involves a video mode change at login...


2022-04-16 10:48:58
Profile WWW

Joined: 2016-06-19 21:33:37
Posts: 337
Location: Columbus area, OH, USA
Reply with quote
I have been using the new 1.29.0.2 this morning with GL worker threading enabled. The texture loading improvement is fantastic. One thing I notice is that with it enabled water appears very dark/black, at least in my set up. Toggling UseGLWorkerFence doesn't alter the condition. Using the 1.28.2.66 release or setting GLWorkerThreads back to 0 seems to return water to its bluer, more reflective state. I've added a screen shot and logs plus my viewer info is:
Code:
Cool VL Viewer v1.29.0.2, 64 bits, Apr 16 2022 09:58:00
Release notes

You are at 277794.0, 255922.8, 21.9  in Muddy Run located at
simhost-02e3361ae4089a138.agni.secondlife.io (34.221.18.58:13012)
Alias: ec2-34-221-18-58.us-west-2.compute.amazonaws.com
Second Life Server 2022-03-24.569934
Release notes

CPU: AMD Ryzen 7 3700X 8-Core Processor (3684.66 MHz)
Memory: 64283MB
OS version: Linux-x86_64 v5.16.19-200.fc35.x86_64 #1 SMP PREEMPT Fri Apr 8 15:34:44 UTC 2022
Memory manager: jemalloc v5.2.1-20220328
Graphics card vendor: NVIDIA Corporation
Graphics card: NVIDIA GeForce GTX 1080 Ti/PCIe/SSE2
OpenGL version: 4.6.0 NVIDIA 510.60.02
Detected VRAM: 11264MB
J2C decoder: OpenJPEG: 1.4.0.635f
Audio driver: OpenAL v1.1 ALSOFT 1.19.1 (OpenAL Soft: OpenAL Soft)
Networking backend: libcurl 7.47.0/OpenSSL 1.0.2u/zlib 1.2.11.zlib-ng
Embedded browser: Dullahan 1.12.4/CEF 100.0.21/Chromium 100.0.4896.88
Packets lost: 0/8330 (0.0%)

Built with: GCC v7.5.0
Compiler-generated maths: SSE2.

Compile flags used for this build:
-O3 -fno-delete-null-pointer-checks -fno-ipa-cp-clone -fno-align-labels -fno-align-loops -fsched-pressure -frename-registers -fweb -fira-hoist-pressure -DNDEBUG -std=c++14 -fno-stack-protector -U_FORTIFY_SOURCE -fno-threadsafe-statics -fPIC -pipe -g -gdwarf -gstrict-dwarf -fno-var-tracking-assignments -fexceptions -fno-strict-aliasing -fvisibility=hidden -fsigned-char -m64 -mfpmath=sse -fno-math-errno -fno-trapping-math -pthread -Wall -Wno-reorder -Werror -DLL_LINUX=1 -DLL_JEMALLOC=1 -DLL_PHMAP=1 -DLL_FMOD=1 -DLL_OPENAL=1 -DXML_STATIC


Attachments:
File comment: GLWorkerThreads == 0
1_29_0_2_regular_water.jpg
1_29_0_2_regular_water.jpg [ 37.55 KiB | Viewed 3429 times ]
CoolVLViewer.zip [23.82 KiB]
Downloaded 120 times
File comment: GLWorkerThreads == -1
1_29_0_2_dark_water.jpg
1_29_0_2_dark_water.jpg [ 49.69 KiB | Viewed 3429 times ]
2022-04-16 13:56:41
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ZaneZimer wrote:
One thing I notice is that with it enabled water appears very dark/black, at least in my set up. Toggling UseGLWorkerFence doesn't alter the condition. Using the 1.28.2.66 release or setting GLWorkerThreads back to 0 seems to return water to its bluer, more reflective state.
What happens if you set "GLWorkerUseForBumpMaps" to FALSE (and relog to make sure water textures will be recreated) ?


2022-04-16 14:29:28
Profile WWW

Joined: 2016-06-19 21:33:37
Posts: 337
Location: Columbus area, OH, USA
Reply with quote
Henri Beauchamp wrote:
What happens if you set "GLWorkerUseForBumpMaps" to FALSE (and relog to make sure water textures will be recreated) ?
Setting that to FALSE and relogging didn't change the appearance. If I toggle RenderTransparentWater, it brightens but is obviously not transparent just like the darker water.


Attachments:
File comment: RenderTransparentWater == FALSE
1_29_0_2_non_transp_water.jpg
1_29_0_2_non_transp_water.jpg [ 39.06 KiB | Viewed 3428 times ]
2022-04-16 14:44:47
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
Well, thing is, even with the future v1.29.0.3 release which allows me to toggle the GL threading on/off on the fly (no more need for a relog), I do not see any difference in water rendering between the two !

I tried with or without ALM, with or without shadows, with minimal or full water reflections and cannot spot any difference in the rendering... I do not see either from where such a difference could come (as long as you do not see any texture corruption, that is).


2022-04-16 15:53:33
Profile WWW

Joined: 2016-06-19 21:33:37
Posts: 337
Location: Columbus area, OH, USA
Reply with quote
Henri Beauchamp wrote:
Well, thing is, even with the future v1.29.0.3 release which allows me to toggle the GL threading on/off on the fly (no more need for a relog), I do not see any difference in water rendering between the two !

I tried with or without ALM, with or without shadows, with minimal or full water reflections and cannot spot any difference in the rendering... I do not see either from where such a difference could come (as long as you do not see any texture corruption, that is).
I have not seen any texture corruptions so far, but I do notice the minimap terrain is often missing so I use the context menu item to refresh. All visible land textures refresh, but the water doesn't 'flash' as if it changed.


2022-04-16 15:56:14
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ZaneZimer wrote:
I have not seen any texture corruptions so far, but I do notice the minimap terrain is often missing so I use the context menu item to refresh.
This is pretty irrelevant: such failures can happen a dozen of times in a raw, and next session you would not encounter any failure... I see this all the time with v1.28.2 as well, during my long drawn out stress-testing sessions with dozens and dozens of sim crossings and TPs...

Quote:
All visible land textures refresh, but the water doesn't 'flash' as if it changed.
The water should not flash either... My terrain reload feature does not reload the water texture (it is not needed by the failing terrain layer composition generation involved in the missing terrain textures issue). :P


2022-04-16 16:05:26
Profile WWW

Joined: 2016-06-19 21:33:37
Posts: 337
Location: Columbus area, OH, USA
Reply with quote
Henri Beauchamp wrote:
The water should not flash either... My terrain reload feature does not reload the water texture (it is not needed by the failing terrain layer composition generation involved in the missing terrain textures issue). :P
Sure. I didn't expect it to really. I have just been trying various things. Clearing cache. Clearing GL shader cache. Other environment settings. I normally run with Parcel. Other settings seem to have the same problem of very opaque water. I'll keep testing. I enjoy the new GL worker threading too much to give it up now. :-) Happy to share my settings file, if you think it would aid in diagnosis. I'm not quite running ultra, nor am I running high. I have a custom set of things, ALM, all shadows, refecltions for avatars and objects, SSAO on, and a few others.
[edit]: I just tried toggling SSAO, as it occurred to me as I was rattling off my customizations. That seems to return things to normal water. Toggling it back on, and the water stays normal.


2022-04-16 16:14:42
Profile

Joined: 2009-03-17 18:42:51
Posts: 5523
Reply with quote
ZaneZimer wrote:
just tried toggling SSAO, as it occurred to me as I was rattling off my customizations. That seems to return things to normal water. Toggling it back on, and the water stays normal.
Please, try reproducing this in LL's performance viewer. I very much doubt it would be caused by anything in my backport, so it should be present as well in LL's viewer; if it is, please report the bug to LL (referencing their own viewer, not mine), on the JIRA.

EDIT: and I just tried with SSAO on or off (including after a viewer restart with it on), and still cannot reproduce your issue...


2022-04-16 16:35:46
Profile WWW
Display posts from previous:  Sort by  
Reply to topic   [ 29 posts ]  Go to page 1, 2, 3  Next

Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group
Designed by ST Software.