The UOC and QUOC Patch - Optimize Firefox (and derivatives) for old hardware!

[looking4awayout]

UPDATE!

I have been doing some other changes to all the versions of the Patch and the Enforcer. Most of those are improvements and refinements of existing features, but my biggest efforts have been focused in the improvement of the way images are rendered in the browser. After messing with the decoding settings, I might have finally found a way to load picture-heavy websites without hammering the CPU too much. On my machine, this allowed me to get an almost lag-free scrolling experience when a picture-heavy webpage is loading.

I also have tweaked the pipelining setting to 128, which again, seems to be better than the previous attempt, when it was set to 150.

As always, please update and let me know how your old computer behaves with this new version, and I apologize for updating the Patch and the Enforcer so often.

Edited October 16, 2019 by looking4awayout

[looking4awayout]

UPDATE!

I have been frantically working on ways to improve the smoothness of the hardware acceleration with the UOC Patch enabled, especially when it comes to scrolling webpages. I have found and disabled another option, scroll anchoring, which seems to mitigate that stuttering issue that affects literally every Mozilla based browser. This, at least on my machine, has mitigated this issue further.

I could use the help of some beta testers, so I can study improvements in a faster way, as the way my machine performs could be very different from the one of another person. If anybody wants to do such, they are always welcome.

Anyway, I have updated the Patch again, for all the versions (45ESR, 38ESR, Macintosh).

P.S. If you have noticed some artifacts during scrolling in New Moon 28/Pale Moon 28 and K-Meleon Goanna, for example the screen flickers in white when you scroll a page with pictures, I can assure it is not caused by the UOC Patch. It is a flaw in the Goanna rendering engine: when tiled compositing -which is essential in order to provide the best acceleration available for a browser- is enabled, for some reason the engine renders the tiles incorrectly when a particularly fast scrolling operation happens.

In the beginning I reported the issue to @roytam1 as I thought it could have been just a bug of New Moon 27 and K-Meleon Goanna, but it's been when I tested the Patch on Pale Moon 28, on my Windows 10 laptop (which I hardly ever use. Everything's done on the RDD, for a proper retro feeling!), that I noticed the same exact issue happening.

So we can safely say that the artifacting during scrolling in Goanna based browsers, is a bug of Goanna itself.

As always, please update to the latest version of the UOC Patch and the Enforcer, which you can find in the main page, and please (again) let me know the outcomes on your computer. I need feedbacks, as they help me to find out if I am doing everything properly or I'm causing havoc to your retro machines.

Happy RDDin'!

Edited October 17, 2019 by looking4awayout

[Mathwiz]

3 hours ago, looking4awayout said:

artifacts during scrolling in New Moon 28/Pale Moon 28

I didn't think you were running version 28. Are you using the UOC Patch with it?

Quote

I tested the Patch on Pale Moon 28, on my Windows 10 laptop

OK, I see now. Just didn't read far enough.

Since you're using "official" PM 28 on an "official" OS, you could report the issue to MCP and see what they say. They may blow it off, but you never know.

Edited October 17, 2019 by Mathwiz

[roytam1]

2 hours ago, Mathwiz said:

Since you're using "official" PM 28 on an "official" OS, you could report the issue to MCP and see what they say. They may blow it off, but you never know.

as they don't recommend people to enable hardware acceleration other than ticking the one in options dialog, your request may be ignored.

[looking4awayout]

Meh, somehow doesn't surprise me. I wonder if it's possible to fix that, unless that bug is deeply rooted in the Goanna engine.

[roytam1]

41 minutes ago, looking4awayout said:

Meh, somehow doesn't surprise me. I wonder if it's possible to fix that, unless that bug is deeply rooted in the Goanna engine.

hard to tell, but for NM27/KM76.2 APZ will be disabled in some cases in recent AF import: https://github.com/roytam1/palemoon27/commit/5510997281dee83f6a7fae37b35fbca99bd4d55e

[looking4awayout]

7 hours ago, roytam1 said:

hard to tell, but for NM27/KM76.2 APZ will be disabled in some cases in recent AF import: https://github.com/roytam1/palemoon27/commit/5510997281dee83f6a7fae37b35fbca99bd4d55e

Yes, I keep APZ disabled on the 38ESR version of the UOC Patch due to that bug in the Goanna engine. On the other hand, I cannot disable tiled compositing, because if I do, performance greatly suffers. So the people who use Goanna based browsers with the UOC Patch must live with the flicker, or else the browser will run in software rendering, and so, crawling on SSE and pre-SSE systems.

[grey_rat]

GF6150

[looking4awayout]

UPDATE!

I have done another major rework on all the versions of the UOC Patch (45 ESR, 38 ESR, Macintosh). The Enforcer has not been touched.

Please update to the latest version and let me know how it runs on your machine!

[looking4awayout]

Yet another update to the UOC Patch for all the versions. Mostly optimizations in the way it manages the connections, as well as disabling some other datareporting features. Please update!

Two of the new changes have been borrowed from the custom prefs.js file made by the user Dronecatcher at the Macrumors forum, which I kindly thank for letting me delve into his configuration file in order to study the potential improvements for the UOC Patch.

Edited October 21, 2019 by looking4awayout

[looking4awayout]

UPDATE!

I have done some major updates to all the versions of the UOC Patch and the 38 ESR Enforcer.

I have reduced the pipelined connections in order to reduce the load on the CPU, as well as reducing the maximum amount of connections per server, which brings benefits not only to performance but also to slow connections such as plain ADSL and dial-up. 

The amount of hardware accelerated layers has been reduced from 20 to 8, providing a noticeable performance increase and one double entry has been removed. 

Please update to the latest version!

 

Edited October 22, 2019 by looking4awayout

[looking4awayout]

NEWS!

I have officially ported the UOC Patch to Firefox 52 ESR based browsers. I had the chance to test the 45 ESR version on the official Firefox 52 ESR and regular Firefox 56, and I was pretty much disappointed by something that I thought would have been taken for granted: tiled compositing and asynchronous pan and zoom after 45 ESR has been broken by Mozilla, and never fixed!

So, I had no other choice than disable those two features, prompting me to release a 52 ESR-and-upwards specific version of the UOC Patch. So now, any browser that uses such a version as a codebase, can be optimized to run better. This new version uses the UOC Enforcer for 45 ESR based browsers, which now is known as the UOC Enforcer for 45 and 52 ESR based browsers.

But bear in mind this important thing:

You will need a 128 bit buss or even better, a 256 bit buss graphics card. I have done the porting of the UOC Patch for 52 ESR based browsers on a Pentium 4 Prescott system fitted with a 64 bit buss graphics card (ATI Radeon X300SE), and the performance increment was negligible at its best, due to the bottleneck caused by the graphics card. So if you want the fastest performance you can achieve with a Mozilla-based browser, I recommend you to go with Roytam1's Firefox 45 ESR SSE, since it has an actual working implementation of tiled compositing and asynchronous pan and zoom (which indeed is enabled by default in the 45 ESR version of the UOC Patch), unless you have a capable graphics card that can fully accelerate the browser even without the offending features.

The newer Firefox builds never cease to amaze me, unfortunately in a negative way.

Edited October 22, 2019 by looking4awayout

[Mathwiz]

<pedantic>

It's "bus." "Buss" is an old English word for "kiss."

</pedantic>

But seriously, thanks for this; I'll work it into my batch file soon (sorry @i430VX).

[i430VX]

53 minutes ago, Mathwiz said:

But seriously, thanks for this; I'll work it into my batch file soon (sorry @i430VX).

It's fine.

But what browsers exactly should utilize this patch?
Should serpent 52 use the FF52 patch or just the usual FF45 version?

[Mathwiz]

Judging from what @looking4awayout put in post 1, the main difference is that tiled compositing and async pan/zoom are disabled in the FF 52 version.

So I'd say NM 28, SM 2.49, BNavigator, and Serpent 52 & 55 should all use the FF 52 patch, since tiled compositing is broken in FF 52 (and so, presumably in all its derivatives too).