RE: cpu not detected(x86)

RE: cpu not detected(x86)

[Petr Vandrovec]

On  7 Aug 01 at 22:09, Dave Jones wrote:
> 
> "The counter is incremented on every processor clock cycle,
>  even when the processor is halted by the HLT instruction or
>  the external STPCLK# pin"

But not if clocks are completely stopped, not through STPCLK, but
just stopped.
 
> "The RDTSC instruction reads the time-stamp counter and is
>   _guaranteed_ to return a monotonically increasing unique value
>  _whenever_ executed, except for 64-bit wraparound".

This only says that two consecutive RDTSC do not return same value.
                                            Best regards,
                                                    Petr Vandrovec
                                                    vandrove@vc.cvut.cz
                                                    

RE: cpu not detected(x86)

[Grover, Andrew]

> From: Nico Schottelius [mailto:nicos@pcsystems.de]
> > I would like to know, if your CPU is on a laptop, are you 
> sure that your
> > power management works well?
> 
> No, seems like ACPI killed the cpu speed.
> I can't undernstand why ACPI is not marked EXPERIMENTAL,
> if those things are normal!

It is.

Regards -- Andy

Re: cpu not detected(x86)

[Alan Cox]

> Longer-term, we need to change the kernel to not use the TSC for udelay, but
> to use the PM Timer, if ACPI is going to be monkeying with CPU power states.

That can be done, and may be a help. 

The TSC timer isnt a very good source on many non intel chips that stop it
to get the best power figures. It also helps with SMP because on an SMP box
the tsc values may not calibrate.

Alan

RE: cpu not detected(x86)

[Grover, Andrew]

> From: Dave Jones [mailto:davej@suse.de]
> Speedstep is voodoo. No-one other than Intel have knowledge of
> how it works. On my P3-700 I've seen speeds range from as low
> as 2MHz[1] -> 266MHz (using an ACPI kernel), and the 550/700 on APM.
> I've also seen other laptops do speed scaling between 2MHz->full clock
> speed whilst on APM.

SpeedStep only drops it to 550 MHz. Any further drops are because of ACPI
processor power or thermal management throwing off your program, because the
current Linux gettimeofday code doesn't think the TSC is ever halted. But,
it is, when the processor is put into C2 or C3. Any benchmark which 1) uses
the TSC and 2) does a sleep() will be wrong.

So, you might try a couple things:

1) Config out the ACPI CPU code. Without it, the system will only exec
"hlt", and the TSC keeps running.

2) Keep the CPU 100% busy throughout the duration of your benchmark.

Longer-term, we need to change the kernel to not use the TSC for udelay, but
to use the PM Timer, if ACPI is going to be monkeying with CPU power states.

Regards -- Andy

PS Your system may also be throttling. It throttles in 12.5% increments, so
that should be borne out in the MHz number if that's what it is doing.


> 
> Run the MHz tester (URL below), and put the box under some load.
> It should increase the MHz accordingly.  How high it goes seems
> to depend on how good your BIOS support for it is.
> 
> Also try switching between ACPI & APM kernels, to see what
> difference it makes.
> 
> regards,
> 
> Dave.
> 
> [1] Actually slower than this, the MHz calculation code takes some
> cycles, so it's an estimate only. http://www.codemonkey.org.uk/MHz.c
> 
> -- 
> | Dave Jones.        http://www.suse.de/~davej
> | SuSE Labs
> 
> -
> To unsubscribe from this list: send the line "unsubscribe 
> linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

RE: cpu not detected(x86)

[Dave Jones]

On Tue, 7 Aug 2001, Grover, Andrew wrote:

> SpeedStep only drops it to 550 MHz. Any further drops are because of ACPI
> processor power or thermal management throwing off your program, because the
> current Linux gettimeofday code doesn't think the TSC is ever halted.

So how come when under heavy load (Ie, when TSC shouldn't be halted),
it only ever reports 266MHz ? Result of overeager CPU throttling?

> But, it is, when the processor is put into C2 or C3. Any benchmark
> which 1) uses the TSC and 2) does a sleep() will be wrong.

24319201.pdf Intel P3 system programmers guide, page 425.

"The counter is incremented on every processor clock cycle,
 even when the processor is halted by the HLT instruction or
 the external STPCLK# pin"

"The RDTSC instruction reads the time-stamp counter and is
  _guaranteed_ to return a monotonically increasing unique value
 _whenever_ executed, except for 64-bit wraparound".

Has this changed ? Or is this the result of a different mechanism?

> Longer-term, we need to change the kernel to not use the TSC for udelay, but
> to use the PM Timer, if ACPI is going to be monkeying with CPU power states.

A worthwhile idea if one is available, given the number of CPUs with
broken TSCs.

> PS Your system may also be throttling. It throttles in 12.5% increments, so
> that should be borne out in the MHz number if that's what it is doing.

If this is the case, it seems to throttle way too aggressively currently,
which could explain the 266MHz limit.

regards,

Dave.

-- 
| Dave Jones.        http://www.suse.de/~davej
| SuSE Labs

cpu not detected(x86)

[Nico Schottelius]

[-- Attachment #1: Type: text/plain, Size: 426 bytes --]

Hello!

I am trying to run 2.4.7 and have heavily problems with my cpu.
The kernel retected another speed at every start! I attached
three times CPUINFO. The cpu in reality is a p3 650 mhz speedstep.
(may switch down to 500 mhz, but 126 _not_).

Who changed something in the 2.4.7 source ?
I am more or less unable to run X with netscape to write this email
with 126 Mhz!

What to do, where to fix ? Please help asap!

Nico



[-- Attachment #2: CPUINFO --]
[-- Type: text/plain, Size: 376 bytes --]

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 8
model name	: Pentium III (Coppermine)
stepping	: 6
cpu MHz		: 161.858
cache size	: 256 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 2
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr sse
bogomips	: 330.95


[-- Attachment #3: CPUINFO2 --]
[-- Type: text/plain, Size: 376 bytes --]

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 8
model name	: Pentium III (Coppermine)
stepping	: 6
cpu MHz		: 127.553
cache size	: 256 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 2
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr sse
bogomips	: 244.94


[-- Attachment #4: CPUINFO3 --]
[-- Type: text/plain, Size: 376 bytes --]

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 8
model name	: Pentium III (Coppermine)
stepping	: 6
cpu MHz		: 162.371
cache size	: 256 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 2
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr sse
bogomips	: 317.84

Re: cpu not detected(x86)

[Dave Jones]

On Tue, 7 Aug 2001, Nico Schottelius wrote:

> Hello!
>
> I am trying to run 2.4.7 and have heavily problems with my cpu.
> The kernel retected another speed at every start! I attached
> three times CPUINFO. The cpu in reality is a p3 650 mhz speedstep.
> (may switch down to 500 mhz, but 126 _not_).

Speedstep is voodoo. No-one other than Intel have knowledge of
how it works. On my P3-700 I've seen speeds range from as low
as 2MHz[1] -> 266MHz (using an ACPI kernel), and the 550/700 on APM.
I've also seen other laptops do speed scaling between 2MHz->full clock
speed whilst on APM.

Run the MHz tester (URL below), and put the box under some load.
It should increase the MHz accordingly.  How high it goes seems
to depend on how good your BIOS support for it is.

Also try switching between ACPI & APM kernels, to see what
difference it makes.

regards,

Dave.

[1] Actually slower than this, the MHz calculation code takes some
cycles, so it's an estimate only. http://www.codemonkey.org.uk/MHz.c

-- 
| Dave Jones.        http://www.suse.de/~davej
| SuSE Labs

Re: cpu not detected(x86)

[Luigi Genoni]

On Tue, 7 Aug 2001, Nico Schottelius wrote:

> Hello!
>
> I am trying to run 2.4.7 and have heavily problems with my cpu.
> The kernel retected another speed at every start! I attached
> three times CPUINFO. The cpu in reality is a p3 650 mhz speedstep.
> (may switch down to 500 mhz, but 126 _not_).
>
> Who changed something in the 2.4.7 source ?
> I am more or less unable to run X with netscape to write this email
> with 126 Mhz!
Sometimes i use an old Pentium 133 classic (no mmx), and it works, well to
do this with X11 and netscape. I think we are starting to be used to PC
that are too powerful. :) If you go to fast how can you see where you are
going? ;)
OK, it was a joke...

I would like to know, if your CPU is on a laptop, are you sure that your
power management works well?

Is it working ok with 2.4.6? and 2.4.8-pre3?

Luigi

Re: cpu not detected(x86)

[Nico Schottelius]

Luigi Genoni wrote:

> On Tue, 7 Aug 2001, Nico Schottelius wrote:
>
> > Hello!
> >
> > I am trying to run 2.4.7 and have heavily problems with my cpu.
> > The kernel retected another speed at every start! I attached
> > three times CPUINFO. The cpu in reality is a p3 650 mhz speedstep.
> > (may switch down to 500 mhz, but 126 _not_).
> >
> > Who changed something in the 2.4.7 source ?
> > I am more or less unable to run X with netscape to write this email
> > with 126 Mhz!
> Sometimes i use an old Pentium 133 classic (no mmx), and it works, well to
> do this with X11 and netscape. I think we are starting to be used to PC
> that are too powerful. :) If you go to fast how can you see where you are
> going? ;)
> OK, it was a joke...
>
> I would like to know, if your CPU is on a laptop, are you sure that your
> power management works well?

No, seems like ACPI killed the cpu speed.
I can't undernstand why ACPI is not marked EXPERIMENTAL,
if those things are normal!

> Is it working ok with 2.4.6? and 2.4.8-pre3?

Sorry, only tested 2.4.7.
Maybe I can test 2.4.8 in the next days.


Nico