Author Topic: OS/4 (technical details only)  (Read 156944 times)

Dave Yeo

  • Hero Member
  • *****
  • Posts: 4788
  • Karma: +99/-1
    • View Profile
Re: OS/4 (technical details only)
« Reply #240 on: March 27, 2021, 06:08:39 pm »
There's an open issue about FFmpeg and AVX at Bitwise's Github. There's also a GCC ticket along the same lines.
The problem is that the object format GCC uses just doesn't allow the right alignments. Ideally would be fixing GCC, not trivial. Another option is not using any AVX C code, also not trivial but doable by looking at trp reports, assembler works fine. Another option might be for the OS4 developers to have a kernel parameter to turn off AVX, which in the short term, might be the simplest.

Ibrahim Hakeem

  • Full Member
  • ***
  • Posts: 120
  • Karma: +5/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #241 on: March 28, 2021, 11:36:28 am »
On my M93P, I've managed to get the latest OS4 kernel to boot, but I get stuck on a loading cursor in the middle of the screen pre XWP starting.

OS4User

  • Sr. Member
  • ****
  • Posts: 406
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #242 on: March 28, 2021, 11:48:23 am »
On my M93P, I've managed to get the latest OS4 kernel to boot, but I get stuck on a loading cursor in the middle of the screen pre XWP starting.

Assuming you have gone through the enclosed HowTo  -  try to reboot a few times, PM Shell or WPS  have some bug which  brings this behavior. I have met it many times on the original kernel too.

Ibrahim Hakeem

  • Full Member
  • ***
  • Posts: 120
  • Karma: +5/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #243 on: March 28, 2021, 02:28:45 pm »
On my M93P, I've managed to get the latest OS4 kernel to boot, but I get stuck on a loading cursor in the middle of the screen pre XWP starting.

Assuming you have gone through the enclosed HowTo  -  try to reboot a few times, PM Shell or WPS  have some bug which  brings this behavior. I have met it many times on the original kernel too.

That turned out to be the solution. I'm aware of this problem but thought there was something else at play after a few reboots didn't do the trick. After a total of 9 reboots, it's finally running properly and is starting every time without troubles now.

Bernhard Pöttinger

  • Jr. Member
  • **
  • Posts: 52
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #244 on: April 04, 2021, 07:38:52 pm »
Hallo,
with IBM Kernel we can switch off cpu cores on boot (acpi.psd /maxcpu=2). Is there an equivalent switch for the os4 kernel.
best regards
Bernhard

OS4User

  • Sr. Member
  • ****
  • Posts: 406
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #245 on: April 04, 2021, 09:01:12 pm »
Hallo,
with IBM Kernel we can switch off cpu cores on boot (acpi.psd /maxcpu=2). Is there an equivalent switch for the os4 kernel.
best regards
Bernhard

what CPU  do you run on ?

Bernhard Pöttinger

  • Jr. Member
  • **
  • Posts: 52
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #246 on: April 05, 2021, 02:26:59 pm »
It's an Intel(R) Core(TM) i7-2760QM CPU @ 2.40GHz - Quad Core CPU. I will try to run with only 2 cores because with 4 cores the the system freezes sometimes; with os4 kernel and ibm kernel. And I think with 2 cores it runs more stable - but with os4 kernel I cannot switch off 2 cores.
best regards

Dariusz Piatkowski

  • Hero Member
  • *****
  • Posts: 1317
  • Karma: +26/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #247 on: April 05, 2021, 02:37:05 pm »
Hi Bernhard,

It's an Intel(R) Core(TM) i7-2760QM CPU @ 2.40GHz - Quad Core CPU. I will try to run with only 2 cores because with 4 cores the the system freezes sometimes; with os4 kernel and ibm kernel. And I think with 2 cores it runs more stable - but with os4 kernel I cannot switch off 2 cores.
best regards

Not sure if this will help or not, but Hobbes has a little utility which allows you to shut OFF individual CPU cores...basically the same thing the built-in SMP 'OS/2 MSP Monitor' does, just uses the much simpler CLI interface.

Here it is => https://hobbes.nmsu.edu/download/pub/os2/util/system/setproc.zip

You could use that to help you troubleshooting...this of course assumes that the issues you are encountering happen only after you successfully boot. If stuff goes wrong during the boot itself setproc may not be able to help.

OS4User

  • Sr. Member
  • ****
  • Posts: 406
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #248 on: April 05, 2021, 08:14:12 pm »
... the system freezes sometimes; with os4 kernel and ibm kernel.

I have freezes sometimes too. May be issue is  in some  drivers  or libs. So far my investigation has not brink a result.

And I think with 2 cores it runs more stable - but with os4 kernel I cannot switch off 2 cores.
best regards

Try the utility suggested by Dariusz.

By the way,   os4apic  does not start HT cores until  switch  /HT is used. May be it helps you.

Bernhard Pöttinger

  • Jr. Member
  • **
  • Posts: 52
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #249 on: April 05, 2021, 08:59:57 pm »
@Dariusz: the suggested tool is only working without acpi. Without acpi the cores stay disabled. With acpi4.bsd (os4krnl) or with acpi.psd (ibm kernel) the cores disabled by setproc.exe will be enabled immediatly.

@OS4User: yes HT cores are disabled until I use the /HT switch; but I cannot disable physical cores.

« Last Edit: April 05, 2021, 09:20:11 pm by Bernhard Pöttinger »

Lars

  • Hero Member
  • *****
  • Posts: 1271
  • Karma: +65/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #250 on: April 05, 2021, 09:24:32 pm »
You have 2 physical cores with 2 logical processors (hyper threads) each. And if hyper threading is not enabled, I'd expect you only see 2 "processors" in OS/2.
This is all IBM marketing bullshit confusion. Works as intended.
Or would Windows show 8 logical processors? If yes, then you would in fact have 4 physical cores.
« Last Edit: April 05, 2021, 09:26:06 pm by Lars »

OS4User

  • Sr. Member
  • ****
  • Posts: 406
  • Karma: +10/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #251 on: April 05, 2021, 09:55:19 pm »
@Dariusz: the suggested tool is only working without acpi. Without acpi the cores stay disabled. With acpi4.bsd (os4krnl) or with acpi.psd (ibm kernel) the cores disabled by setproc.exe will be enabled immediatly.

ACPIDAEMON.EXE  switches them ON

Dariusz Piatkowski

  • Hero Member
  • *****
  • Posts: 1317
  • Karma: +26/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #252 on: April 05, 2021, 10:06:13 pm »
Bernhard,

@Dariusz: the suggested tool is only working without acpi. Without acpi the cores stay disabled. With acpi4.bsd (os4krnl) or with acpi.psd (ibm kernel) the cores disabled by setproc.exe will be enabled immediatly.

So a couple of things I need to point out, I should have actually done so to start off with.

1) hyperthreading
Per the official ACPI read me, the OS/2 drivers do not support hyperthreading...the '/CPUMASK' is meant to shut off hyperthreading in case the actual system BIOS does NOT give you the ability to do so. Since we are talking OS4 here I'm thinking this is not applicable, but for what it's worth I thought I should mention it.

2) setproc
As it turns out given the last couple of releases of the ACPI drivers, when used with the normal OS2 kernel and the ACPI Power Daemon, there is indeed a problem with forcing any particular core into an OFF state. In my case (desktop hardware config, not a laptop) I actually disable the Power Manager feature of the ACPI Daemon, so the Power Manager does not work at all.

If you do not disable the Power Manager function the ACPI Daemon will force the core into it's own chosing of either OFF or ON state, most likely ignoring the selection you had made. This makes sense but happens to be a non-viable configuration for my needs, so Power Manager is permanently shut off and given that my hardware is ON 24x7 that's just fine.

Dave Yeo

  • Hero Member
  • *****
  • Posts: 4788
  • Karma: +99/-1
    • View Profile
Re: OS/4 (technical details only)
« Reply #253 on: April 05, 2021, 10:41:51 pm »
You have 2 physical cores with 2 logical processors (hyper threads) each. And if hyper threading is not enabled, I'd expect you only see 2 "processors" in OS/2.
This is all IBM marketing bullshit confusion. Works as intended.
Or would Windows show 8 logical processors? If yes, then you would in fact have 4 physical cores.

It's an I7, until recently they had 4 real cores and another 4 logical if HT is turned on. Intel marketing BS, not IBM. For quite a while it was pretty consistent, I3=2 core, I5=4 core, I7= 4+4 core with HT. More other features with the higher numbers too. Lately it has become more confusing with Intel competing  with AMD on number of cores.

Lars

  • Hero Member
  • *****
  • Posts: 1271
  • Karma: +65/-0
    • View Profile
Re: OS/4 (technical details only)
« Reply #254 on: April 06, 2021, 03:09:24 am »
You have 2 physical cores with 2 logical processors (hyper threads) each. And if hyper threading is not enabled, I'd expect you only see 2 "processors" in OS/2.
This is all IBM marketing bullshit confusion. Works as intended.
Or would Windows show 8 logical processors? If yes, then you would in fact have 4 physical cores.

It's an I7, until recently they had 4 real cores and another 4 logical if HT is turned on. Intel marketing BS, not IBM. For quite a while it was pretty consistent, I3=2 core, I5=4 core, I7= 4+4 core with HT. More other features with the higher numbers too. Lately it has become more confusing with Intel competing  with AMD on number of cores.

Sorry, meant to say Intel marketing BS. But if Windows shows 8 logical processors, then yes, it's 4 real cores.

One of the problems with ACPI.PSD and /CPUMASK is, that the core numbering is not necessarily consistent/logical (I don't know why this is so).
Therefore, if you think that /CPUMASK=0x55 will skip one hyper thread per core, then this is not guaranteed. It's trial and error. And therefore, the only reliable solution is to turn off Hyperthreading via BIOS (which not all BIOSES support).