Cpustress - incorrect detection of cpus and keyboard problem

[bob3248]

In UBCD 5.2.0 and 5.2.1 cpustress does not correctly detect the number of cpus. I tested it on a desktop Intel Core 2 Quad Q9650 (4 cores, 4 threads) and a notebook Intel Core i5-3210M (2 cores, 4 threads) and it both cases it detects only 1 cpu and runs only 1 instance of the mersenne prime test.

Also, on the desktop system the keyboard works in the UBCD menus but stops working as soon as I start cpustress. The keyboard is a Microsoft Natural Keyboard Pro connected with USB. On the notebook the keyboard worked normally.

[Explorer09]

The number of CPUs in the script is from the /proc/cpuinfo file.

I can help you debug if you can give me the contents of /proc/cpuinfo of your computer.

Can you get one in Parted Magic and upload it?

[ady]

The PMagic version included by default in UBCD is i586-compatible. The dual x32/x64 bits PMagic version might give a different result.

Fortunately, version 5.2.x of UBCD simplifies the update of the PMagic version in a customized UBCD. To update (or to use a different version), _some_ of the files in "/pmagic/" need to be updated and the old "pmagic*.SQF" file needs to be deleted. The "

/pmagic/boot/syslinux*.*cfg*

" file(s) should be tweaked too, but it is not a complicated process (and it doesn't affect CPUStress in particular).

[bob3248]

Here is the cpuinfo file from the Core 2 Quad Q9650 desktop. I got it from the 32 bit version of Parted Magic 2013_02_28 (Not the Parted Magic in UBCD)

[Explorer09]

I actually found where the problem is, and I think Ady and Victor should be responsible to this.
The i586-compatible kernel from PMagic does not support SMP. Therefore when I boot from that kernel only one CPU core can be used. I don't know whether that improves hardware compatibility or not, but I'm sure it does break CPUstress.

[ady]

@Explorero09,

I was the one that reminded Victor that the standard PMagic dual 32/64 bits requires so many resources that it can't boot in some systems (like some Celeron ones that don't support PAE, for example). Also the necessary amount of RAM (and storage space) for PMagic keeps rapidly increasing too, but that's not an issue for this topic.

Although mentioning or suggesting something / anything to be used in official releases of UBCD is not the same as being responsible for it, I wouldn't mind "taking responsibility" for it . BTW, I don't regret that suggestion, and if it were up to me I wouldn't change Victor's decision regarding this.

Now, in a more serious tone, I already posted above that the solution is to customize UBCD with a different variant of PMagic, or with any appropriate kernel for the hardware in question.

There is no simple way to cover all the requirements with one official release of UBCD. In fact, PMagic is slowly turning into a more generic distro, and IMO at some point UBCD should be released with a script to add whichever PMagic version the user would choose to download by himself, with no PMagic included by default in the official UBCD releases (only the script).

FWIW, there are ways to combine different versions / variants of PMagic into the same media if necessary, but that's off-topic here.

[Explorer09]

I was the one that reminded Victor that the standard PMagic dual 32/64 bits requires so many resources that it can't boot in some systems (like some Celeron ones that don't support PAE, for example). Also the necessary amount of RAM (and storage space) for PMagic keeps rapidly increasing too, but that's not an issue for this topic.

I see, but I myself don't see that customization is a solution.

Because of the change of the kernel, the tools in CPUstress won't be able to stress multiple CPUs. I consider this a feature loss, and I would need to recover that feature.

For now, I have only 3 choices:
1. Use the i686 PAE+SMP kernel from PMagic.
2. Keep using the i586 kernel and forget about the multiple CPUs.
3. Build a kernel from scratch.

Which one should I take? Please suggest one for me.

[ady]

@Explorer09,

I hope this is not considered too much "off-topic".

The following is just my opinion and nothing else: each user should customize UBCD / CPUStress and use whichever kernel see fit for each situation.

One way to provide more options would be to have CPUStress (initrd + append boot options) together with several pmagic kernels:
_ bzimage_i586(x32-nonSMP-nonPAE)
_ bzimage_i686(x32-SMP-PAE)
_ bzimage64

Since PMagic keeps updating the kernel between its frequent releases, this would be one way (for users) to keep CPUStress tests updated and covering different hardware (old and new).

UBCD releases are much less frequent than PMagic and than new hardware (CPUs). The utility of having PMagic, bootable on a broader range of hardware, surpasses the general utility of hardware burn-in tests.

This is the reason why I see fit, for the official (and relatively infrequent) releases of UBCD, to use PMagic_i586. In fact, as I already mentioned, my opinion is that no version of PMagic should be included but a script only, reducing the size of UBCD to a third and "forcing" the user to customize it according to his needs. Running simple scripts already included in UBCD is not difficult, and any potential user that attempts to use any of the programs included in UBCD should be able to execute them. If a user is not capable of running the scripts, then he should be better not using UBCD at all.

Since PMagic is frequently updated, the customization with newer PMagic releases in whichever variant (i586 or dual x32/x64) is the best path, IMHO. A clear help text regarding the usage of CPUStress with different kernels (supporting different hardware) could be helpful for final users.

To be clear, in the current releases of UBCD, there is no ready-made script to update PMagic. I hope this changes in future releases (but such interesting and important discussion is off-topic here).

[Explorer09]

@ady:
I would prefer good out-of-the-box support in the kernel. I will try to compile one on my own.
The other 3 PMagic kernels will be left for customization.