• Welcome to OS2World OLD-STATIC-BACKUP Forum.
 

News:

This is an old OS2World backup forum for reference only. IT IS READ ONLY!!!

If you need help with OS/2 - eComStation visit http://www.os2world.com/forum

Main Menu
Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - rwklein

#1
Quote from: Blonde Guy on 2012.09.29, 22:31:08
Some computers will still need throttle control. So I hope it remains in the toolkit, so we can implement throttling to keep those computers cool. HLT is great, but sometimes it isn't enough. In your example above, IPIHLT = 0. Doesn't that mean that HLT isn't working on that computer?

Thats from my T60. And just about any x86 cpu suports HLT instructions:

http://en.wikipedia.org/wiki/HLT

Roderick Klein
Mensys
#2
Quote from: StefanZ on 2012.09.29, 15:45:06
Hello Roderick,

thank you for your reply.

The main idea behind my question was if there is anything specific what I can do - as a single end user - to make sure that my system is better support in the future versions of ACPI?
...or is it better to wait for newer versions of ACPI that will get progressively better?

(I decided to act because it seems a bit ridiculous why my (standard Intel) system should not be supported at any time by any  of (even  the old) versions of acpi. I'm getting worried a bit that my system might actually never be fully supported...).

Thanks.
St.

You conclude from the fact that the throttle support is not supported that your chipset is not fully supported basically.
Before eCS 2.1 we never shipped any type of CPU method to cool down the CPU.

People had noted that the old _W4 kernel would keep CPU's cool. The SMP kernel we use in eCS does did have that HLT support the _W4 kernel has! So powerman.exe demon was implemented. It turned out these two ways it tries todo throttle support depend on the BIOS implementation. A lot of CPU's did this not work.

Infact based on the measurements David (who works on ACPI). The two throttle ways powerman.exe supports
slow down your system as it has to get from a "speed" up to a full speed.

Over the years we looked at implementing a way to get HLT support on an SMP system. But it was not easy as the SMP kernel has different ways of doing stuff then _W4 kernel does. The HLT support does not have the negative performance effect the throttle support powerman.exe offers. On top of that the HLT support provides even better power savings then the technology that powerman.exe provides. So the GUI to setup powerman and powerman.exe itself will go away in eCS 2.2.

The HLT support is effective automaticly by installing recent ACPI versions.
You can see it under acpistat.exe

ACPI name [CPU0]
IPIGenCount = 27585      IPICount    = 84680
IPIHLT      = 0          IdleCount   = 213371     BusyCount   = 25384

This works on just about any system with an AMD/Intel CPU. The success rate is WAAAY higher then with the ways powerman.exe supports.

So from that perspective it seems your throttle support we already support your system optimal. As an added bonus if the system is not being used other CPU cores are switched of automatically by ACPI and switched on back within a blink of an eye! So when it comes to throttle support of the CPU we are in good shape!

Roderick Klein
Mensys
#3
Quote from: StefanZ on 2012.09.28, 19:27:05
What should one do, if running on Intel ICH9 family chipset and gets this out of ACPI:

This system does not support CPU Performance Control via ACPI.

This system does not support CPU Throttling.

Thermal information:
  No Thermal information is available for this computer


Does it mean my system is so nonstandard?  :'(
St.

The latest versions if ACPI the use the HLT instruction with the SMP kernel (its active automaticly) just like it was in _W4 kernel. The old throttle stuff is less efficient and the GUI will be removed. Since the HLT instruction is much more efficient
and works on just about any CPU.

Roderick Klein

#4
Quote from: melf on 2012.09.26, 23:24:57
Well, the graphical marks as far as I can see seem to be extremely low. On my Thinkpad T43 the PM Marks is 118 with Panorama (compared to  454 with SNAP) which is really quite good. It has been some talk about bad MTRR values and Panorama (frankly I don't know what that is, but it seem to be able to badly affect Panorama performance, but should be possible to adjust I think).

MTRR is the magic word. And a lot of systems on which SNAP runs in VESA mode and is slow its the same thing.

Roderick
#5
Quote from: miturbide on 2012.09.15, 01:19:41
I'm having some issues with ACPI on a Thinkpad L420 (7856-3WS).
Mainboard: Intel HM65 (Cougar Point)
Processor: Intel Core i5-2410M

After installing ACPI, it boots fine, but when I try to go to the "ACPI Power Manager" settings it says:
"The Power Manager service is not available. Either ACPIDaemon is not running or your system does not support power management"

It didn't work with or without /VBE. The Standy by goes to sleep but does not turn on back.

By the way, I don't use eCS 2.1 on this machine running on real hardware (only virtualized) since the trackpoint and touchpad does not work (only the external mouse works) and the Intel Centrino 6205 wireless is not supported.

Its really the question if we are going to keep the old throttle tool alive (powerman). I will discuss this with David but from what I understand the HLT instruction with the new ACPI versions works without any trouble on most systems. And the support powerman offers is slower and from what I understood David has measured just has negative effect on the performance of a system and does not even conserve as much energy as the HLT instruction (that ACPI activates automatically just like the _W4 kernel did).

As for your trackpoint I send you a driver you could try.

Roderick Klein
Mensys
#6
Quote from: DougB on 2012.09.14, 22:46:56
QuoteI just tried this release of ACPI on the Thinkpad R51 (1838-PS0) and the standy mode is working.

My Lenovo ThinkPad T510 will go to sleep, but it will not turn the screen back on, after waking it up (the back light stays off). Otherwise, S/R appears to work (although with no activity on the screen, it is really difficult to tell if it really did wake up). Wireless never did work, so that is waiting for the Multimac project to catch up.


Well try /VBE behind the PSD=ACPI.PSD second if it does not work. After suspend see if you can ping the machine from another machine. And you can check if caps lock works (the led turns on).

Telnet does not work. It seems the video card must be back alive for telnet to work over the network.

Roderick Klein
Mensys
#7
Sometimes it is bad luck. But a bug has been found in APM.ADD by Joachim and me from Mensys.
David tested the 3.21.01 build on systems at Warpstock and on his own 10 computers at home,
Suspend resume did not work.

It turned out the old developer left a bug in the APM.ADD thunking code. The bug has been reported and fixed within 48 hours by David. We are now testing 3.21.02 and hopefully will release this ASAP.

Roderick Klein
Mensys
#8
Quote from: danielnez1 on 2012.07.28, 01:46:10
The current version locks up my Compaq Presario CQ57 almost immediately  :-[

Hello Daniel,

I hope you will also (if you have not do so already) open a ticket here:
http://svn.netlabs.org/acpi#SubmittingTickets

We do not have access to all equipement. So feedback is needed.

Please note the ticketing system does not send back an email when the ticket is updated.
So you need to check from time to time.

Thanks,

Roderick Klein
Mensys

P.S. I know its pain the Netlabs ticketing system does not send an email when a ticket is updated.
#9
Quote from: eirik on 2012.07.24, 02:45:34
Great job by the ACPI team, although some functionality is still not there: primarily FnFx buttons on my Lenovo T410.  It now suspends properly, but the screen does not come alive (I'll post a separate msg. on the track where it belongs, but I just want to applaud the ACPI team for their progress.

Keep up the good work!  More and more of us has come to realize the complexity in ACPI, but also its value in terms of making modern hardware available to eCS users.

The goal has always been to make ACPI easy. I especially think with the work David has done the boot success rate has gone up lot. The new HLT support in ACPI package 3.20.03 seems a good step forward to keeping your CPU cool.

Suspend resume will still need more work. Try on your system the /VBE switch behind the PSD=ACPI.PSD line. It might bring your screen back to live.

Roderick Klein
Mensys
#10
Quote from: Sigurd on 2012.06.27, 22:36:22
Hi,

yes, thanks, but it showed up as an interrupt problem. LAN WLAN USB3 - all three use IRQ16, deactivating LAN Driver does not help. I think I have to take a closer look at how to redirect IRQ with ACPI.

See picture.


As far as I know the new ACPI versions from David don't support reroute of an IRQ anymore. This feature either does not work anymore or is at least not recommended to be used anymore. I understood from David this IRQ re-route with ACPI is something that is not really with in line of the standards.

While we where told in the past that multiple devices on a single IRQ are not possible I have started to belief this is not true.
Both David and Steve Levine mentioned that this should not be a problem. And if there is a problem then something in a driver is going wrong.

In other words most IRQ re-route strategies introduced in the past where a workaround for the true cause of other problems.
As for SCSI cards that don't work in APIC mode that, for that we recommend you use the /VW switch. Which brings your system back to 15 IRQ's.

Roderick Klein
Mensys
#11
Quote from: miturbide on 2012.07.01, 01:14:34
Hi Roderick

QuoteBut thats the whole thing we (Mensys) is allowed to open source the code of the DDK.

I didn't understand well this phrase. Do you mean that Mensys can create a derivate work with the IBM DDK code and turn it open source?
(Please notice that Open Source mean, releasing the software under any of the approved open source license by OSI www.opensource.org/licenses)

If so, can Mensys make the Lars USB drivers work to be open sourced?   Lars told me that he can not open source it (it is on a restrictive access SVN on netlabs) because the IBM DDK license does not allows it.

Regards


I told you this before Martin. The DDK code can not be open sourced. The IBM license does not allow this. Thats also the reason ACPI is closed source because it relies on the DDK from IBM. It would need to be required that somebody rewrites the DDK from scratch I don't consider this to happen any time soon.

What is possible is there is a serious developer that has a good possibility writing this driver that the developer "works" for Mensys.
That would be possible under the DDK license. Mensys once downloaded the DDK and as company can have employees work on the code.

But somebody just can not out of the blue get access to the code.

Roderick
#12
Quote from: DougB on 2012.06.30, 21:49:38
QuoteThis has been on the Mensys project list for over 3 years now. But from my point of view it not that useful since the base USB stack still needs more work.

I agree that the base USB stack needs more work, however, it is now becoming urgent to get a USB 3.0 driver. The simple fact is, that there are some machines that are shipping with hardware that needs the USB 3.0 driver, and they don't have any other options (other than adding another card to support USB). Over the next year, or so, I expect to see no other options.

I have not followed this dossier. But are there mainboards out there that no longer have controllers on board that work with USB 2.0 drivers ? That they add USB 3.0 controllers that the first. But USB 3.0 devices can, as far as I can tell be hooked up to a USB 2.0 port. 

Roderick Klein
Mensys
#13
Quote from: miturbide on 2012.06.30, 19:51:18
Roderick

Maybe it is a good time to invest on putting out the driver development knowledge to the public. The blasting truth is "there is not enought people with the knowledge to create OS2-eCs drivers", but it will be good to try to find somekind of solution.

Where can we get that knowledge of the internal workings of the OS/2 USB stack? Can I post it to the EDM/2 Wiki under Creative Common?

Well there are enough people around to write OS/2 device drivers. I would need to dig run past my list of names again. But does people you also need to able to pay them (that means not a to high charge) and they must have enough time to write the code.

And then these people also need to be good enough. I'm not going to open up the dossier on how a certain developer worked out for  us (Mensys). Its also in general rare to find people for any OS that really write decent device driver code!

About where to get documentation on the inner workings of the USB stack. Its for 99% in the DDK code.
And understanding does interfaces is possibly not easy, I don't know. But thats the whole thing we (Mensys) is allowed to open source the code of the DDK. But if you think you have a developer that can write such a driver we can grand him access.

From one of the Netlabs developer workshops there where 2 presentations on the USB stack. I think you can find them on ecomstation.ru. In general if you look at the revision history of the USB drivers from Lars it also goes to show to some extend that the USB protocol is also a kind of trial and error game. Its not supposed to be like that but just like with ACPI the standards in real hardware do not add up the way they are documented.

Roderick Klein
Mensys
#14
While suspend resume we don't support officially its still being worked and moving forward.
More and more systems are starting to work and 3.20.03 version has a new Intel ACPI code in the PSD that has fixes for suspend resume for none Windows systems (if I understood David his explanation correctly).

While I see that Melf his system does not suspend resume correctly anymore with the current version the suspend/resume is a very complex procedure. But as we do have (as far as it now seems) a good working HLT support that seems to be working on a lot of systems!

The SMP kernel never supported this. We have in the test lab also other systems see other work much better with suspend resume then ever before. While not ready for production usage never have so many systems worked with suspend resume.

To give you an idea of how systems are working now. System that never would even go to sleep now suspend and then resume and you can access the system via the kernel debugger (via the serial cable).  But things go wrong in restoring the video buffer.
Thats why with testing suspend/resume try the /VBE switch.

As time will move forward more progress is being made that will only see the success rate of suspend/resume go up.
We will never have the coverage that Windows has. But that is not that weird in it self.  Suspend resume from my point of view is screwed standard. With that I mean that a standard OEM install of a Windows 7 CD will also not work on all systems.
As most hardware vendors preload drivers for there own equipment.

So right now suspend resume is officialy not supported as more stuff is being debugged. But progress is being made.

Hope the new ACPI versions will help keep your systems cool and reduce your electricity bill :-)

Roderick Klein
Mensys
#15
Quote from: Andi710 on 2012.06.24, 10:08:08
http://www.os2world.com/content/view/21818/2/
QuotePosted by Martin Iturbide - Saturday, 23 June 2012


I have jus. .. .

Runs well here on a Phenom II X6. However, with SNAP video enabling 4 cores is still the limit. (6 cores only work with plain VGA, SNAP hangs at WPS boot-up)

And the reactor temperature widget for XCenter does not show anything except of "Zzz".
Does anybody have a working version that can show system/CPU temperature as measured using ACPI?

Cheers
Andreas

So what happens if you run with Panorama do the 6 cores work ? If that works then we know SNAP has an SMP defect!
We have somebody working on SNAP at the moment. Based on your report I opened a ticket:

http://svn.ecomstation.nl/snap/ticket/21

Regards,

Roderick Klein
Mensys