OS/2 works slower in SMP mode if compare with 1 CPU

[Eugene Gorbunoff]

This topic is important:
ACPI daemon, PowerMan, WinSendMsg -- https://forum.os2world.com/index.php?topic=4124.0

Seems that OS/2 works slower if you run in MultiCore mode, SMP mode, with ACPI daemon enabled.
(IF compare with MultiCore disabled, running with 1 CPU)

OF course, you feel this if running heavy applications: web-browser, packing files, video/music play, ..

So, it's necessary continue the research.

[Roderick Klein]

This topic is important:
ACPI daemon, PowerMan, WinSendMsg -- https://forum.os2world.com/index.php?topic=4124.0

Seems that OS/2 works slower if you run in MultiCore mode, SMP mode, with ACPI daemon enabled.
(IF compare with MultiCore disabled, running with 1 CPU)

OF course, you feel this if running heavy applications: web-browser, packing files, video/music play, ..

So, it's necessary continue the research.

First point: Make certain you run the latest ACPI. Updates where made to the ACPIdeamon.

Second point: How do you measure the performance ? If possible turn off hyperthreading.

Roderick

[Dariusz Piatkowski]

Hi Eugene!

This topic is important:
ACPI daemon, PowerMan, WinSendMsg -- https://forum.os2world.com/index.php?topic=4124.0

Seems that OS/2 works slower if you run in MultiCore mode, SMP mode, with ACPI daemon enabled.
(IF compare with MultiCore disabled, running with 1 CPU)...

So I'm curious: what prompted you to bring this topic back up for discussion?

Not minding at all here, in fact seeing as our previous thread had me testing ACPI 3.23.18 and now I'm on 3.23.19, I went back to re-test some of this. In doing so I found that the '/NOHLT' parameter (which I had left in my CONFIG.SYS) doesn't appear to improve things anymore, in fact, it being present there almost causes the previous slow-response issues to be exhibited again!!! (which is quite the opposite result of what I saw with 3.23.18 version of ACPI drivers)

The 3.23.19 change log shows the following:

3.23.19 - 18-Feb-2025 - David Azarewicz

Reduced memory requirements.
Power manager changed to be less agressive enabling secondary cpus.
Updated to ACPICA 20241212. See https://acpica.org/ for details.

Perhaps those changes are what has fixed/addressed the previous issues we were seeing?