Quote
roytam1

Quote
siria
IMO if some users really want those 7% more speed, they can always upgrade. Until then, they have no major probs, everything works.

But on the other hand, if the big majority of users have massive freeze probs, this is an absolute killer-bug.

For playing L5.1 with ref-frames > 3, it is not only 7% improvement with assembly/SSE2 enabled.

YouTube limits for ref-frames from 1 to 3, even for 4K clips. Tested experimentally.
I admit, third-party services can have a different setting.

Updated:
Conducted real tests, I encoded video with 16 ref-frames.
Difference is increasing consumption of memory.

360p fullscreen (first with ref-frames = 4, second with ref-frames = 16)


720p fullscreen (first with ref-frames = 4, second with ref-frames = 16)


1080p fullscreen (first with ref-frames = 4, second with ref-frames = 16)



Tests conducted on a one core of Core2Quad (3 GHz).



Edited 13 time(s). Last edit at 01/04/2018 04:27AM by hermes.

Comparing LAV editions (without any instructions and with SSE2) on a very weak netbook
(Intel Atom 1.6 GHz, dual core) we have another picture.

Intel Atom fully supports SSE2 and with a weak processor this saves from overload.

Here advantage SSE2 up to ~25% (for ref-frames = 16).



Tested on 720p video. Win for mobile platforms is obvious.


Trying to watch 1080p (for netbook it's difficult)


Video with ref-frames more than 4 is not common.



Edited 11 time(s). Last edit at 01/04/2018 07:14AM by hermes.

Some more experiments ¬_¬. Resource consumption while playing video on Core2Quad
(3 GHz, all cores enabled), 8 GB RAM, Server 2003 DC SP2 x86 as OS (PAE is ON).

The system consumes about 256 MB of RAM. For browser I made RAM disk = 128 MB.
(tested with SSE2 version, for maximum effectiveness)

720p



For scale, only one process (only seems what here ~1.15%, really ~3-4%).



1080p



4K



Responsiveness of browser resembles a work with notepad . Maybe video for tests and were successful, but nevertheless, optimization of browser is beyond praise. It is just perfect!



Edited 1 time(s). Last edit at 01/04/2018 07:01AM by hermes.

Quote
rodocop

And I plan ask roytam1 for some possible changes in old modules keeping the same idea in mind: make KM better and more user-friendly.

"User-friendly" - this term can be interpreted very broadly.
Any chance for disclosing what those changes should accomplish?

Wouldn't the Improvement requests forum be best suited for such tasks?



Edited 1 time(s). Last edit at 01/04/2018 10:23AM by Yogi.

Quote
Yogi

Quote
rodocop
And I plan ask roytam1 for some possible changes in old modules keeping the same idea in mind: make KM better and more user-friendly.

"User-friendly" - this term can be interpreted very broadly.
Any chance for disclosing what those changes should accomplish?
Wouldn't the Improvement requests forum be best suited for such tasks?

IMHO user-friendly and improvement requests are two very different concepts.
If you follow all suggestions made by all users you will get this:


The first requirement for user-friendly is something useful for mostly all people, that works fine along a good period of time with no crashes. I agree @rodocop words, because IMHO stability, robustness and confidence about you are using is the best mouth to mouth marketing. A few months ago I tested Vivaldi browser and after four or five crashes in a week I gave up definitely. I have never turn back to test it again and I won't recommend to use it. Never. From 1000 suggestions probably 1 will be acceptable. I always have KM76RC2pro by @Rodocop in an USB key and I carry it with the security that it will work in all computers I will find, everywhere. That's user friendly for me.
:cool:



Edited 1 time(s). Last edit at 01/04/2018 07:13PM by J.G..

What's the application you're using to get those program-specific performance graphs?

Process Explorer.

But I consciously use old version 9.23.0.0 (where charts more contrast) and it can run literally everywhere: from Windows95 to Windows Server 2016 (i.e. supports both Ansi and Unicode systems in variation x86 / x64, with monitoring for native x64 processes). With new version, with use on systems from Vista+, can display GPU loading (for me it is not required). But run on older OS for these new versions is not possible (Win9x and to me it is critical).
So choice is very wide.

Actual version (v.16.21) https://download.sysinternals.com/files/ProcessExplorer.zip

More older version:
http://www.oldversion.com/windows/process-explorer/
or
http://www.oldapps.com/process_explorer.php

Of course, I can upload my version (if need to), there are also older versions (with their advantages and disadvantages, for Win3.11 + Win32s : ) https://en.wikipedia.org/wiki/Win32s .



Edited 2 time(s). Last edit at 01/06/2018 03:38PM by hermes.

Just curious, or may be i get all this wrong.
I copied all versions one by one to test. But nothing changed on this youtube HTML5 page. I mean H264 and MSE & H.264 support, always red exclamation mark, see the screenshot:



Wasnt these dll's supposed to enable MSE & H.264 on youtube?

@WasiAzeem

On fully default, latest Goanna (20180825)
http://kmeleonbrowser.org/forum/read.php?19,143253,147321#msg-147321
fully support all tests on youtube.com/html5 without any changes of browser settings.



Edited 1 time(s). Last edit at 09/19/2018 03:16PM by hermes.

oh sorry, I was using K-Meleon76RC2.

yes indeed KM-Goanna passes all of the tests.

I was trying to fix KM 76 RC2 that showed those exclamation marks.

It appears that those dll's are not enabling KM 76 RC2 for H264 stuff. However, KM 76 RC2 has no problem on youtube, it only shows red excamation marks against MSE & H.264 and H.264 on https://www.youtube.com/html5.

KM-Goanna is amazing, but I also want KM 76 RC2 to be more compatible.

Thanks