ACPI vs Standard PC HAL's

[bob661]

I've been using the Standard PC Hal on my computers for quite a while for stability purposes. But it seems that newer hardware may require the ACPI HAL to be used. This especially true on laptops. What do you guys think?

 

[bob661]

Ah crap! This should be in the OS forum. Can someone move this?

 

[bob661]

Bump for a move?

 

[Peter]

"Standard PC" for stability? Hahahahaha ... that's just as good as the "IRQ conflict" mythology.

Scratch that. You won't get full performance out of a modern PC when using a legacy HAL. You might even lack features - Cool&Quiet, Speedstep, APIC interrupt controllers, and a couple of smaller details don't work unless you're running ACPI.

(Of course, if your board's BIOS has a buggy ACPI implementation, then you'll experience problems. But this is 2005, everyone should have their act together by now.)

 

[ProviaFan]

The last time I had to use Standard PC was on an AMD K6-2 400MHz system due to a buggy ALI chipset. ACPI is much better for anything made in the past few years, 99.999995% of the time.

 

[imported_vdorta]

My new MSI RS480 didn't completely shut down and I traced the problem down to a Standard PC HAL. A XP SP2 update fixed that by updating to the correct "ACPI Single Processor HAL" or something like that.

 

[flibidyflob]

Originally posted by: bob661
I've been using the Standard PC Hal on my computers for quite a while for stability purposes.

That's an interesting notion. I second Peter's response.

 

[bob661]

Thanks very much for the replys guys!

 

[bob661]

Here's a question from a friend of mine:

I have used the ACPI kernel and here is the problem. ACPI uses 1 interrupt to pool and EMULATE many IRQS using memory slot chunks to address that IRQ. If you have many devices using 1 IRQ, like Video, Sound, Raid & SATA Controllers, Firewire, Ethernet, SCSI, USB ....etc and are using them simutaneously, like video, sound, ether, usb and raid and its aggressive enough YOU WILL have lock ups because too many devices are trying to use that 1 IRQ that has been emulated as many. All motherboards including todays only have 15 IRQS and no more. The ACPI takes one of those and shares it as many. This is how they get around the issue. It is a poor resolution to the issue. If you believe there is anymore than 15 hardware IRQs than you don't know the X86 architecture.

 

[ProviaFan]

Originally posted by: bob661
Here's a question from a friend of mine:

I have used the ACPI kernel and here is the problem. ACPI uses 1 interrupt to pool and EMULATE many IRQS using memory slot chunks to address that IRQ. If you have many devices using 1 IRQ, like Video, Sound, Raid & SATA Controllers, Firewire, Ethernet, SCSI, USB ....etc and are using them simutaneously, like video, sound, ether, usb and raid and its aggressive enough YOU WILL have lock ups because too many devices are trying to use that 1 IRQ that has been emulated as many. All motherboards including todays only have 15 IRQS and no more. The ACPI takes one of those and shares it as many. This is how they get around the issue. It is a poor resolution to the issue. If you believe there is anymore than 15 hardware IRQs than you don't know the X86 architecture.

Since that is an absolute statement, I can say plainly that it is absolutely wrong. To have said a few years ago that on rare occasions badly written device drivers or poorly designed hardware might not behave well and thus cause trouble with shared IRQs would have been true, but to make such a statement as the above in this day and age is nothing but nonsense!

 

[bob661]

The reply from my friend:

ProviaFan, you need to look futher than the end of your nose when reading like the words that say "and its aggressive enough YOU WILL have lock ups". Todays ACPI maybe a little more advanced than the first BUT the issue of only 15 physical hardware IRQs still exist and no matter how you or anybody else trys to justify it that this is the truth. Until hardware manufacturers redesign thier motherboards to have 28 to 32 hardware IRQs you will be at the mercy of the hardware and the crappy ACPI. Live with it!

 

[ProviaFan]

Originally posted by: bob661
The reply from my friend:

ProviaFan, you need to look futher than the end of your nose when reading like the words that say "and its aggressive enough YOU WILL have lock ups". Todays ACPI maybe a little more advanced than the first BUT the issue of only 15 physical hardware IRQs still exist and no matter how you or anybody else trys to justify it that this is the truth. Until hardware manufacturers redesign thier motherboards to have 28 to 32 hardware IRQs you will be at the mercy of the hardware and the crappy ACPI. Live with it!

My reply to him: you need to quit getting stuck in computing tradition and folklore. I have seen some fairly loaded systems (in terms of hardware), and have yet to encounter ANY of these supposed problems that you speak of. Obviously, even if there might be some "issue" in theory, the vast majority of the time, it is simply not a problem at all in practice.

 

[Green Man]

Originally posted by: bob661
Here's a question from a friend of mine:

I have used the ACPI kernel and here is the problem. ACPI uses 1 interrupt to pool and EMULATE many IRQS using memory slot chunks to address that IRQ. If you have many devices using 1 IRQ, like Video, Sound, Raid & SATA Controllers, Firewire, Ethernet, SCSI, USB ....etc and are using them simutaneously, like video, sound, ether, usb and raid and its aggressive enough YOU WILL have lock ups because too many devices are trying to use that 1 IRQ that has been emulated as many. All motherboards including todays only have 15 IRQS and no more. The ACPI takes one of those and shares it as many. This is how they get around the issue. It is a poor resolution to the issue. If you believe there is anymore than 15 hardware IRQs than you don't know the X86 architecture.

You need to evolve your thinking beyond ten years ago. The chained 8295 PICs of 16 IRQs and DOS have been replaced by the Multiprocessor APIC. By current specifications, there may be up to 8 I/O APICs in a system each supporting 24 or more interrupts. All single processor motherboards I have seen recently support 24 interrupts, and the only reason you're still only seing 16 is because you haven't enabled ACPI.

If you have a single processor system, it is recommended that you enable your APIC because it allows faster IRQ handling. If you have a multiprocessor system, it is required. The Only reason to force the APIC to revert to legacy 8259 mode is if you are running Win9x, because APIC isn't supported.

 

[Peter]

Exactly. You're limited to 15 hardware IRQ lines as long as you're running on the legacy pair of PICs. With the APICs up and running (by means of either MPS or ACPI HAL), the number of interrupt controllers, and thereby the number of hardware interrupt input lines, is limited to 254.

Actual implementations typically have 20 to 24 lines for simple, single-APIC systems as found in notebooks and desktops, and twice or even more for servers and workstations. Your average Opteron or Xeon board typically uses two or three APICs with 16 to 24 inputs each.

"IRQ sharing" problems are an urban myth too - possible only on ISA hardware. PCI or newer, this no exist. If your problem feels like it is an IRQ sharing problem, then what you actually have is a buggy driver. OK, it's a well rooted myth - I've had $15k halfway-round-the-world trips to customers to prove this.

bob, your writing isn't only utterly wrong, it's actually quite hilarious in its made up technobabble. "Memory slot chunks" ... that's nearly Star Trek quality nonsense

 

[bob661]

Peter,
Those comments were from my friend. Not I. I just want information about this subject.

 

[Peter]

I see. You may go kick your friend now