Hey all,

I'm new here and I have seen a couple of requests for PAE enabled 32bit kernels but have not seen whether this has been implemented yet.

I started using Liquorix kernels and just love the performance on my laptop, however, I prefer to use 32bit as I seem to get better performance with Games and even some applications. Especially with Flash and some other necessary tools only supported in 32bit.

My laptop has 6gigs of ram and I like to be able to to use most of it for virtualization and other tools that require lots of ram, but mostly play games and love the eye candy Linux has :).

Is it possible to have a PAE enabled kernel option for us to download? I am not familiar with how to build kernels in Ubuntu, but have done it in Arch and Gentoo, so with the proper instructions, I would not mind attempting to configure my own with PAE support, but would rather have something already supported.

Thanks for you time and hope you all had a great Thanksgiving!

Joe
Back to top

You won't lose anything by switching to 64bit. I'm running Debian Sid in 64bit right now and I'm still able to play all my games that only have a 32bit build, wine included. All that's necessary is to install the 32bit libraries which your distribution should supply in their repositories.

Flash also has a native 64-bit plugin. I can't provide you any numbers to compare versus their 32-bit plugin, but it's definitely not slow. I watch 1080p videos on youtube in full screen; maybe that says enough.

It could also be that I have a fast system and that any performance regression due to using a 64-bit kernel is unmeasurable for me:



You can read the specs and come to your own conclusions, but I won't be building a PAE 32-bit kernel when most difficulties of running a 64-bit desktop actually from the distribution not supplying the proper 32-bit compatibility libraries to make a 64-bit desktop practical.
Back to top

I don't believe there is a current flash-plugin for 64 bit, adobe some time ago pulled all their 64 bit stuff, promising to restore it later, probably needed to do some core rewrites or something.

I can't agree with damentz however that you will see no difference, in especially the non free arena, you will see some significantly poorer support for 64 bit, that's been the case now for as long as I can remember, it has gotten much better, but it's still an issue that pops up routinely enough for me to not even remotely consider 64 bit for my free desktop.

I do agree with damentz however in not wanting to provide a pae kernel, 3.5 gigabytes is totally fine for almost anything normal users will ever do, with a tiny fringe outside of that who might actually use more, almost always because of using multiple virtual machines at the same time, that's the only real time I can think of you get close to 3+ gig of ram usage in the real world.
Back to top

Thanks for the reply, damentz, techAdmin!

Here is my system...



As you can see, its not that old either ;).

I understand if you do not want to support an additional package, but how would it be if you simply create the config for 32bit to include 64gb memory in your base package?

This way, the user will have access to all their memory if they have it?

I also have to disagree with you on the merits of 64 bit computing. There is just not enough reason to move to 64bit on the desktop that compares to 32bit. Sure, I have used 64bit for 3 to 4 years and while there are 32bit libs available for 64bit, the system has to cswitch a ton to 32bit code to get there and that takes memory and processor performance. Why emulate 32 bit when you can just run it ;).

Now, for general desktop use, this might be negligible for most, but when you are a performance nut and expect your system to run 100% for gaming or whatever, you do notice a difference :).

Besides, 64bit is just not as mainstream for desktops. Servers are a different story, but desktops just don't need that much power as of yet.

This is my opinion, anyway. I do appreciate you making a high performance kernel. Its been a while sense I have been excited to run one; last time was when Con Kolivas had the CK tree :). Boy, those were he days, hu?

Anyway, I have compiled my own kernel with PAE support and even added some bells and whistles while I was at it.

Thanks again, and best wishes!

Joe
Back to top

Sorry for insisting, but I work with the popularization of Linux in Brazil since 2004.

Our country, as well as several other countries also have several cities that use dial-up.

The dial-up drivers only work on 32-bit kernel.

I think the addition of a compilation "PAE" will be little extra work for you, but will facilitate my work and that of many who are trying to strengthen the use of Free Software in countries like Brazil.
Back to top

Debian already has linux-image-2.6.32-5-686-bigmem

which should be fine for most users who want to run absurdly large amounts of RAM they don't need to do anything real in almost all cases except to look at it and note that it is there and used.

bigmem enables > 3.5 gig of ram for 32 bit.

I believe there's a performance drawback with that if I remember right, which is probably why liquorix will not use it.
Back to top

techAdmin, the idea is not to use "PAE" on the kernel by default liquorix.

But included in the package for Debian build another option, that offers a feature called "PAE". Available to the user two kernel versions, one with "PAE" and one without.

On another topic also briefed a small change needed in Plymouth package that allows the use of kernel liquorix usually in Ubuntu distribution.

With these changes I believe will be useful to liquorix many more users than at the time.
Back to top

bigbruno, it's up to damentz to decide to maintain separate branches that do not in any way benefit him personally.

He's already done a very good job with his past/future/main branches, expecting even more to be added to that for almost no gain in the real world I feel is not very realistic, nor would it really be something most people would use or be interested in.

I think you are massively overestimating the number of users who would benefit from this, my guess is that the numbers in the real world would be in the tens. Not hundreds, tens.

Linux users are prone to vastly overestimating the user bases of their systems, by a factor often of about 10, at least.
Back to top

Please see this thread for liquorix based PAE kernel.

this is also, by the way, the correct solution, someone makes it who wants it, and then distributes it, thus leaving damentz free to engage in the stuff that actually interests him.
Back to top