Hare you saying that a Slackware recompiled with i686 optimization will have the same performance?

x86_64 here.

Interesting poll, but I suspect the results are skewed in favour of x86_64 and the newer processors. I imagine that there are many x86_64 users who drop by linuxquestions quite frequently (to catch up on flashplayer and other compatibility issues etc) and would see the poll. I can also imagine that there are i486 workhorses out there whose owners have less reason to keep up with the latest news, and dare I say it, don't participate regularly on this forum.

tobyl

No, I am saying that it is pointless for Slackware to add another 32-bit version of the distro. It adds nothing of value and only takes away precious time in maintaining it.

Eric

IMHO, compiling Slackware for i686 instead of i486 makes complete sense. It would improve performance a bit for i686 users, an architecture supported by most computers out there, and i[45]86 users would still have older Slackware releases that are still maintained. However, it's simply not the time to do it. It could have been done a few years ago, but now it makes no sense. Today we're sold x86_64 machines and an x86_64 version of Slackware exists, which improves performance much further. If you want performance and taking full advantage of your architecture, use that version. If performance is not your worry, the difference between i486 and i686 is so small that the advantage of knowing the distribution will run on almost any 32-bit machine you could find today surpasses the benefits of rebuilding the system, even if at incremental steps, for i686. Not worth the effort. If Slackware64 didn't exist, things would be different, but it's not the case anymore.

IMO it is rather pointless to bother. It is just as rg3 says, if you want performance you should be running Slackware for x86_64.

If you need Wine to play World of Warcraft, but refuse to go multilib on Slackware64, then just install Slackware 32-bit. You will certainly have a PC that is beefy enough that you do not have to be bothered with performance gains when march=i686 is applied.

Read http://lists.centos.org/pipermail/ce...ne/006063.html which has a nice performance breakdown for all the different values of march and mtune. This bit is interesting:

And that is exactly why 32-bit Slackware is going to remain as-is.

Eric

But to get the top performance, you might then want special Slackware distributions compiled with -march=nocona or -march=core2, depending on whether your 64-bit intel processor is a p4 or something newer. In addition, you would need three different distributions for 64-bit AMD cpus. Hmm...

Yes, that is true. The performance difference is probably larger between 32-bit and 64-bit than between i486 and i686.

As for optimizing per processor, that is unlikely, and I doubt it will increase performance by much.

If you care that much about performance, install Gentoo not Slackware.

I'd rather retain support for i586, mainly because I have an AMD Geode which is i586:I also have a desktop with x86_64 running -current.

Thanks for showing how pointless the exercise is :-)
Maintaining even more versions of the same distro is not going to pay for Pat's daily bread. Sales of Slackware is not going to rise but the required effort will double.

Eric

I dont think anyone ever suggested maintaining two 32bit distributions.

I generally compile my own glibc which is i686-optimized
 

and usually set properly optimized cflags in my slackbuilds, not to mention that I compile my own kernels, which I usually distribute to relatives -some of them quite capable to compile a kernel, but too lazy to do it. So in the end we have a Slackware-based distro at home. (I changed the penguin logo with the Slack logo... couldn't resist... )

This is why I love Slackware! Let's face it -- I don't notice any speed difference in actually using apps when testing out other i686 distros than with stock x86 Slack. A "slightly" faster boot and ten-second faster shutdown (on Ubuntu & Trisquel) doesn't impress me at all. And look at what you have to go through to get that. I mean...ech.

There's nothing (other than knowledge) stopping anyone from recompiling for i686. It will preserve Pat&Company's precious resource of Time and everybody Gets What They Want without having to go all LFS.

As far as antique hardware...my dad used my original 1983 IBM PC (with the 8088 proc and 64k ram) to do his monthly rental house invoices as recently as late 2007. Never replaced the cmos battery (yet) or anything -- which is really amazing.

(I can just imagine some guy with a VAX 11/730 in his basement waiting to top this post.)

It does DNS and SMTP. It was also my router and firewall for a long time, but it can't keep up with network throughput anymore. Plus, my internet connection is 14Mb, and the only two ISA NICs I have are only 10Mb.

The hard drive is backed up, and it's on its third or fourth drive. The BIOS can't even address 20% of the drive, but it doesn't really need to. The original drive that the system was installed on back in 1999 was only 512MB. Otherwise, it's all 1990s hardware.

It will NEVER run any newer software, except when Sendmail and BIND patches come out, or if somebody ever were to patch the 2.2 kernel again.

Both my desktop and laptop (not my netbook) use AMD Athlon CPUs, 1.2GHz and 1.5GHZ respectively. Both run Slackware 13.1. I know they're old but they still suit my needs and thanks to Slackware and Fluxbox I never feel pressured to upgrade.

I use 2 machines, a pentium III with 933 GHz with Slackware 13.1and the other one AMD ATHLON 2.8 GHz with Slackware 64 13.1

Old bones with less old bones ... LoL