Jump to content

My Browser Builds (Part 3)


Recommended Posts


 Well, that's nice to hear. Anyway, since changes have been made to the D3D9 initialization code(Browser will now try to create a pure device, if supported), we need someone with a good GPU to test this new build, to check if everything is working correctly. I do have a GF9600GT laying around, but i can't it use right now, due to power supply constraints(My current PSU can't provide enough power to it and rest of the components).

Edited by XPRTM
Link to comment
Share on other sites

54 minutes ago, XPRTM said:

 Well, that's nice to hear. Anyway, since changes have been made to the D3D9 initialization code(Browser will now try to create a pure device, if supported), we need someone with a good GPU to test this new build, to check if everything is working correctly. I do have a GF9600GT laying around, but i can't it use right now, due to power supply constraints(My current PSU can't provide enough power to it and rest of the components).

With GTX260 everything works. But I don't know how to find out this hardware or software vertex shader

Link to comment
Share on other sites

15 hours ago, XPRTM said:

 Well, that's nice to hear. Anyway, since changes have been made to the D3D9 initialization code(Browser will now try to create a pure device, if supported), we need someone with a good GPU to test this new build, to check if everything is working correctly. I do have a GF9600GT laying around, but i can't it use right now, due to power supply constraints(My current PSU can't provide enough power to it and rest of the components).

quote from Sami:

Quote

as I again try to recall, to keep it short:

950 does not fully support T&L (blending) - using D3DCREATE_SOFTWARE_VERTEXPROCESSING is pointless for acceleration as it is just emulating GPU features using CPU (cf. WARP), which can be done better using BasicCompositor that is optimized for this task. In a matter of fact only D3DCREATE_HARDWARE_VERTEXPROCESSING should be used, as other options are meant for either debugging or providing fallback (which in this case is already provided).

 

Also as I quickly try to recall: PUREDEVICE cannot be used for initialization as Get calls are used in the code. It may be used for DXVA though.

@XPRTM what do you think?

Link to comment
Share on other sites

2 hours ago, roytam1 said:

quote from Sami:

@XPRTM what do you think?

  Hardware T&L has nothing to do with blending, they are two separate things. To clarify what the new code does, now the device capabilities are checked, so if the GPU supports hardware vertex processing(like grey-rat's GTX260), it will be used. Software vertex processing is only used if the GPU doesn't support hardware vertex processing, like the GMA3000.

  Now even with software vertex processing, there is still hardware acceleration for scene rasterization(that's where blending may be used), granted performance may not be good, but this was allowed in the older pre SM 3.0 D3D9 spec. Just read the D3D9 docs, and you will see. I don't  know why he mentioned WARP, since that is for D3D10.1+ only, and is a pure software solution, and not comparable to rasterization HAL devices.  

 Regarding the pure device remark, if Get calls are used like he said, then that part of the code must be changed, otherwise it may cause problems. Just give me some time, i will fix it.

 

Edit:

I have sent you a PM with the corrected code.

Edited by XPRTM
Link to comment
Share on other sites

We have fully hardware pixel shaders. GMA900-3150 do not support Direct2D or DX10+ or DXVA

Run Firefox 43 on winXP with GMA and enable WebGL Angle. This is many faster than full soft WebGL LLVM

https://habrastorage.org/webt/xu/md/3e/xumd3evotbrsc0qdeg4ro4qm4uw.jpeg

https://habrastorage.org/webt/9b/vk/gy/9bvkgyzufmfyozjprdsks7_toyg.jpeg

roytam1,  is it possible to reduce the restriction of 3.0 shaders to 2.0 to return the hardware Webgl Angle for ATI Radeon 9500-X1250, NVIDIA FX5 *** and Intel GMA 900-3150?

 

 

Edited by grey_rat
Link to comment
Share on other sites

^That's a funny issue. I get this on official Pale Moon 31.2.0.1 if I have the developer tools open when loading that page, but works otherwise. If console is open, one JS file says:

Unexpected exception sending payload. Ex:[object DOMException]{}

If I connect to it via Proxomitron (just pass-through), then no error and it loads normally, regardless of whether the developer tools are open or not...

Edited by UCyborg
Link to comment
Share on other sites

On 10/1/2022 at 9:39 PM, John Carter said:

Discord site crashes every time I log in. Happens on Basilisk/Serpent too. Can this be fixed?

Already reported upstream:
https://forum.palemoon.org/viewtopic.php?p=232312#p232312

It's the same underlying issue I spoke of in my Aug 29th post, i.e. missing Regex Unicode Property Escapes support in UXP ...
As to the "whether it can be fixed?", my linked post has relevant info...
German developer of palefill (martok) has recently fixed the Google Drive breakage, so perhaps all is not lost -
post in palefill's issue tracker (GitHub account required) ...

But if I were you, I'd use web-discord inside a Chromium-based browser (it's what they target, after all...); at the start of this year I used to frequent a Discord server and when I tried my default Vista browser, Serpent 52.9.0 (Discord "worked" there at the time), the whole experience was very sub-optimal (increased sluggishness all around the GUI :angry: ... ); for Discord, I had to switch to 360EEv12/v13 (360EEv11 had some glitches, as in not "scrolling" certain areas of the GUI :( ... ) .

PS: Logging-in to Discord is such a PITA, though :realmad: ... I have an internet connection where my IP address changes every now and then (dynamic/not dedicated IP); Discord then "needs" to re-authorise my new IP, meaning: I have to first supply my account credentials to them, then an IP-authorisation e-mail is sent in my inbox (with a short-lived link), I need to click that link in the same browser I'm trying to log-in, get the confirmation message my new IP has been "authorised", then attempt to log-in a second time (a form of 2FA, as I lack a mobile phone) ... One of the (many) reasons I avoid "social networks" like the plague :angry: !

Edited by VistaLover
Link to comment
Share on other sites

13 hours ago, XPRTM said:

  Hardware T&L has nothing to do with blending, they are two separate things. To clarify what the new code does, now the device capabilities are checked, so if the GPU supports hardware vertex processing(like grey-rat's GTX260), it will be used. Software vertex processing is only used if the GPU doesn't support hardware vertex processing, like the GMA3000.

  Now even with software vertex processing, there is still hardware acceleration for scene rasterization(that's where blending may be used), granted performance may not be good, but this was allowed in the older pre SM 3.0 D3D9 spec. Just read the D3D9 docs, and you will see. I don't  know why he mentioned WARP, since that is for D3D10.1+ only, and is a pure software solution, and not comparable to rasterization HAL devices.  

 Regarding the pure device remark, if Get calls are used like he said, then that part of the code must be changed, otherwise it may cause problems. Just give me some time, i will fix it.

 

Edit:

I have sent you a PM with the corrected code.

got another mail from Sami:

Quote

I allow myself to give some further explanation and commentary on GPU acceleration issue.

Firstly, she mentioned WARP because that is also an software emulation like with D3DCREATE_SOFTWARE_VERTEXPROCESSING. (- ‿. )

By blending I meant overall composition. Compositor depends on vertex support (see vertex buffer code) and to my understanding only hardware processing mode (and mixed in hardware mode) will utilize GPU. Worth to mention is that basic drawing is already GPU accelerated.
In my humble opinion, in context of discussion on bug 637024, it may be worth considering doing some benchmarks before using software fallback by default.

Nonetheless, I think it is great that you are attempting to restore hardware acceleration.

Lastly canvas poisoning is still broken (PoisonValue result is not used).

 

Link to comment
Share on other sites

11 hours ago, roytam1 said:

got another mail from Sami:

 

I wouldn't call software vertex processing emulation. Generally, it is just lighting and position calculations(or vertex shaders) being executed on the CPU FPU, instead of the GPU. All pixel processing(pixel shaders, blend ops, etc) are still done on the GPU hardware. Vertex buffers are obviously available(and usable) on software vertex processing devices, with the only caveat being that it must be stored on system memory(for quick CPU access), but this is irrelevant for the older GMA's, since they have no embedded video memory anyway, and just use system memory as video memory instead.

  Like i said previously, with this new device initialization code, the GPU capabilities are checked, and hardware vertex processing will be used if the GPU supports it. If it does not, then software vertex processing is used. Nothing is being forced on anybody.

Edited by XPRTM
typo.
Link to comment
Share on other sites

19 hours ago, AstroSkipper said:

The login to the website https://outlook.live.com/ (formerly Hotmail) doesn't work anymore in New Moon 28 (2022-09-29). Tested with a clean profile, too. Here is a screenshot:

NM28-Outlook-Login.png

I don't know how long this problem has existed. :no: Haven't logged in there for a while. :dubbio:

16 hours ago, UCyborg said:

^That's a funny issue. I get this on official Pale Moon 31.2.0.1 if I have the developer tools open when loading that page, but works otherwise. If console is open, one JS file says:

Unexpected exception sending payload. Ex:[object DOMException]{}

If I connect to it via Proxomitron (just pass-through), then no error and it loads normally, regardless of whether the developer tools are open or not...

It's really a strange, "funny" issue. If I turn the website style off and then on again, the fields for the login are visible again. Therefore, with switching the website style I finally could log into my account. I wonder seriously what's going on there. In the past, I never had such problems. :dubbio:

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...