Re: cpufreq on Pentium M

Re: cpufreq on Pentium M

[Matthew Miller]

In linux.kernel, Dave Jones wrote:
> For info folks, I've now had a dozen or so mails with the same cache
> descriptors.  Please don't send me any more centrino based outputs,
> I have plenty now, thanks 8-)

Hi Dave. This is the output from my Celeron 600A system -- it's a tiny bit
different from what others posted to the list (an 0x86 instead of an 0x87).
My understanding is that this CPU has half the L2 cache of the "real"
Pentium M. Hope this is helpful and not annoying. Also, a question: I had
assumed that the lack of info in /proc/cpuinfo was simply that an
informational problem, and that the cache is actually working. Am I
mistaken? (I.e. will having the kernel support this be a huge performance
increase?) Thanks!



/proc/cpuinfo:

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 9
model name	: Mobile Intel(R) Celeron(R) processor     600MHz
stepping	: 5
cpu MHz		: 595.408
cache size	: 0 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 mce cx8 sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 tm
bogomips	: 1186.20


x86info -c:

x86info v1.9.  Dave Jones 2001, 2002
Feedback to <davej@suse.de>.

Found 1 CPU
Family: 6 Model: 9 Stepping: 5 Type: 0 [ Original OEM]
unknown TLB/cache descriptor:
	0xb0
unknown TLB/cache descriptor:
	0xb3
Instruction TLB: 4MB pages, fully associative, 2 entries
unknown TLB/cache descriptor:
	0x86
unknown TLB/cache descriptor:
	0x30
Data TLB: 4MB pages, 4-way associative, 8 entries
unknown TLB/cache descriptor:
	0x2c



-- 
Matthew Miller           mattdm@mattdm.org        <http://www.mattdm.org/>
Boston University Linux      ------>                <http://linux.bu.edu/>

Re: cpufreq on Pentium M

[Dave Jones]

On Thu, Jun 12, 2003 at 10:45:49PM -0400, Matthew Miller wrote:

 > Hi Dave. This is the output from my Celeron 600A system -- it's a tiny bit
 > different from what others posted to the list (an 0x86 instead of an 0x87).
 > My understanding is that this CPU has half the L2 cache of the "real"
 > Pentium M. Hope this is helpful and not annoying.

Yeah, I'd realised that was also missing. I'll get it in sync with 2.5
after Marcelo makes a 2.4.21

 > Also, a question: I had
 > assumed that the lack of info in /proc/cpuinfo was simply that an
 > informational problem, and that the cache is actually working. Am I
 > mistaken? (I.e. will having the kernel support this be a huge performance
 > increase?) Thanks!

performance-wise, it may make a small difference on SMP systems, as it
affects the balancing of the scheduler. On UP, shouldn't be any difference.

		Dave

Re: cpufreq on Pentium M

[Matthew Miller]

On Fri, Jun 13, 2003 at 11:52:24AM +0100, Dave Jones wrote:
>  > mistaken? (I.e. will having the kernel support this be a huge performance
>  > increase?) Thanks!
> performance-wise, it may make a small difference on SMP systems, as it
> affects the balancing of the scheduler. On UP, shouldn't be any difference.

Ok, thanks. It's not an SMP subnotebook. :)

-- 
Matthew Miller           mattdm@mattdm.org        <http://www.mattdm.org/>
Boston University Linux      ------>                <http://linux.bu.edu/>

cpufreq on Pentium M

[John Goerzen]

Hello,

I am running on a Thinkpad T40p laptop, which has a 1.6GHz Intel
Pentium M CPU (this is their "Centrino" CPU; *NOT* the same thing as
the Pentium 4 M).

I have tried 2.4.21-rc7-ac1, but it reports:

cpufreq: Intel(R) SpeedStep(TM) for this chipset not (yet) available.

Yet my BIOS lets me configure SpeedStep for it by name.  Are there any
plans in the works for this?

While we're at it, I'm concerned that Linux is ignoring the sizable
cache available on this platform:

$ cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 9
model name      : Intel(R) Pentium(R) M processor 1600MHz
stepping        : 5
cpu MHz         : 1598.686
cache size      : 0 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 mce cx8 sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 tm
bogomips        : 3191.60

Re: cpufreq on Pentium M

[Dave Jones]

On Wed, Jun 11, 2003 at 11:13:26PM -0500, John Goerzen wrote:

 > I am running on a Thinkpad T40p laptop, which has a 1.6GHz Intel
 > Pentium M CPU (this is their "Centrino" CPU; *NOT* the same thing as
 > the Pentium 4 M).

Stay tuned. Jeremy Fitzhardinge wrote a driver for centrino style
speedstep. It's currently getting the kinks worked out on the cpufreq list.
It should turn up in 2.5 sometime real soon, and at some point, maybe
someone will backport it.

 > While we're at it, I'm concerned that Linux is ignoring the sizable
 > cache available on this platform:
 > 
 > $ cat /proc/cpuinfo
 > processor       : 0
 > vendor_id       : GenuineIntel
 > cpu family      : 6
 > model           : 9
 > model name      : Intel(R) Pentium(R) M processor 1600MHz
 > stepping        : 5
 > cpu MHz         : 1598.686
 > cache size      : 0 KB

Looks like missing cache descriptors. Grab x86info[1] and mail me
the output of x86info -c

		Dave

[1] http://www.codemonkey.org.uk/x86info/

Re: cpufreq on Pentium M

[Martin List-Petersen]

Citat Dave Jones <davej@codemonkey.org.uk>:

> On Wed, Jun 11, 2003 at 11:13:26PM -0500, John Goerzen wrote:
> 
>  > I am running on a Thinkpad T40p laptop, which has a 1.6GHz Intel
>  > Pentium M CPU (this is their "Centrino" CPU; *NOT* the same thing as
>  > the Pentium 4 M).
> 
> Stay tuned. Jeremy Fitzhardinge wrote a driver for centrino style
> speedstep. It's currently getting the kinks worked out on the cpufreq list.
> It should turn up in 2.5 sometime real soon, and at some point, maybe
> someone will backport it.
> 
>  > While we're at it, I'm concerned that Linux is ignoring the sizable
>  > cache available on this platform:
>  > 
>  > $ cat /proc/cpuinfo
>  > processor       : 0
>  > vendor_id       : GenuineIntel
>  > cpu family      : 6
>  > model           : 9
>  > model name      : Intel(R) Pentium(R) M processor 1600MHz
>  > stepping        : 5
>  > cpu MHz         : 1598.686
>  > cache size      : 0 KB
> 
> Looks like missing cache descriptors. Grab x86info[1] and mail me
> the output of x86info -c
> 

Got the same thing here on a Dell Latitude D600

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 9
model name	: Intel(R) Pentium(R) M processor 1600MHz
stepping	: 5
cpu MHz		: 1594.855
cache size	: 0 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 mce cx8 sep mtrr pge mca cmov pat clflush dts
acpi mmx fxsr sse sse2 tm
bogomips	: 3185.04

x86info will follow (just want to upgrade my kernel first, the thing is on
2.4.21-rc2-ac2 now)

Regards,
Martin List-Petersen
martin at list-petersen dot dk
--
Talkers are no good doers.
		-- William Shakespeare, "Henry VI"

Re: cpufreq on Pentium M

[Mark Watts]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Thought this may help:
- From a Dell D600 w/ 1.6Ghz Pentium M cpu...

Mark.


cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 9
model name      : Intel(R) Pentium(R) M processor 1600MHz
stepping        : 5
cpu MHz         : 1598.671
cache size      : 0 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 mce cx8 sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 tm
bogomips        : 3191.60




x86info v1.11.  Dave Jones 2001, 2002
Feedback to <davej@suse.de>.

Found 1 CPU
eax in: 0x00000000, eax = 00000002 ebx = 756e6547 ecx = 6c65746e edx = 49656e69
eax in: 0x00000001, eax = 00000695 ebx = 00000816 ecx = 00000180 edx = a7e9f9bf
eax in: 0x00000002, eax = 02b3b001 ebx = 00000000 ecx = 00000000 edx = 2c043087

eax in: 0x80000000, eax = 80000004 ebx = 00000000 ecx = 00000000 edx = 00000000
eax in: 0x80000001, eax = 00000000 ebx = 00000000 ecx = 00000000 edx = 00000000
eax in: 0x80000002, eax = 20202020 ebx = 20202020 ecx = 65746e49 edx = 2952286c
eax in: 0x80000003, eax = 6e655020 ebx = 6d756974 ecx = 20295228 edx = 7270204d
eax in: 0x80000004, eax = 7365636f ebx = 20726f73 ecx = 30303631 edx = 007a484d

Family: 6 Model: 9 Stepping: 5 Type: 0
CPU Model: Unknown CPU Original OEM
Feature flags:
        Onboard FPU
        Virtual Mode Extensions
        Debugging Extensions
        Page Size Extensions
        Time Stamp Counter
        Model-Specific Registers
        Machine Check Architecture
        CMPXCHG8 instruction
        SYSENTER/SYSEXIT
        Memory Type Range Registers
        Page Global Enable
        Machine Check Architecture
        CMOV instruction
        Page Attribute Table
        CLFLUSH instruction
        Debug Trace Store
        ACPI via MSR
        MMX support
        FXSAVE and FXRESTORE instructions
        SSE support
        SSE2 support
        Automatic clock Control
        Pending Break Enable


unknown TLB/cache descriptor:
        0xb0
unknown TLB/cache descriptor:
        0xb3
Instruction TLB: 4MB pages, fully associative, 2 entries
unknown TLB/cache descriptor:
        0x87
unknown TLB/cache descriptor:
        0x30
Data TLB: 4MB pages, 4-way associative, 8 entries
unknown TLB/cache descriptor:
        0x2c

/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device

MTRR registers:
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
/dev/cpu/0/msr: No such device
MTRRcap (0xfe): MTRRphysBase0 (0x200): MTRRphysMask0 (0x201): MTRRphysBase1 (0x202): MTRRphysMask1 (0x203): MTRRphysBase2 (0x204): MTRRphysMask2 (0x205): MTRRphysBase3 (0x206): MTRRphysMask3 (0x207): MTRRphysBase4 (0x208): MTRRphysMask4 (0x209): MTRRphysBase5 (0x20a): MTRRphysMask5 (0x20b): MTRRphysBase6 (0x20c): MTRRphysMask6 (0x20d): MTRRphysBase7 (0x20e): MTRRphysMask7 (0x20f): MTRRfix64K_00000 (0x250): MTRRfix16K_80000 (0x258): MTRRfix16K_A0000 (0x259): MTRRfix4K_C8000 (0x269): MTRRfix4K_D0000 0x26a: MTRRfix4K_D8000 0x26b: MTRRfix4K_E0000 0x26c: MTRRfix4K_E8000 0x26d: MTRRfix4K_F0000 0x26e: MTRRfix4K_F8000 0x26f: MTRRdefType (0x2ff):

1598.66 MHz processor (estimate).




- -- 
Mark Watts
Systems Engineer
QinetiQ TIM
St Andrews Road, Malvern
GPG Public Key ID: 455420ED

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+6Im5Bn4EFUVUIO0RApBZAKCMhyv/g89z/3q86HgHyVCMbK19jwCgr6d7
ADZKclFpLmbiyvhHsrpNBCI=
=XSHX
-----END PGP SIGNATURE-----

Re: cpufreq on Pentium M

[Dave Jones]

On Thu, Jun 12, 2003 at 03:10:01PM +0100, Mark Watts wrote:

 > Thought this may help:
 > - From a Dell D600 w/ 1.6Ghz Pentium M cpu...

Yep, the latest bunch of cache descriptors are missing from 2.4
(They're in 2.5 already). I did send these to Marcelo, but they
seem to have got lost when we entered -rc stage.
I'll resend them for 2.4.22pre

 > 
 > unknown TLB/cache descriptor:
 >         0xb0
 > unknown TLB/cache descriptor:
 >         0xb3

uninteresting (tlb sizes)

 > Instruction TLB: 4MB pages, fully associative, 2 entries
 > unknown TLB/cache descriptor:
 >         0x87

1MB l2 unified cache.

 > unknown TLB/cache descriptor:
 >         0x30

32KB L1 I cache

 > Data TLB: 4MB pages, 4-way associative, 8 entries
 > unknown TLB/cache descriptor:
 >         0x2c

32KB L1 D cache

Current CVS version of x86info already supports all these too btw,
(just in case someone was tempted to hack a patch for that against 1.11).
I'll do another release sometime this weekend.

		Dave

Re: cpufreq on Pentium M

[John Goerzen]

On Thu, Jun 12, 2003 at 03:44:50PM +0100, Dave Jones wrote:
> On Thu, Jun 12, 2003 at 03:10:01PM +0100, Mark Watts wrote:
> 
>  > Thought this may help:
>  > - From a Dell D600 w/ 1.6Ghz Pentium M cpu...
> 
> Yep, the latest bunch of cache descriptors are missing from 2.4
> (They're in 2.5 already). I did send these to Marcelo, but they
> seem to have got lost when we entered -rc stage.
> I'll resend them for 2.4.22pre

Is it already too late for 2.4.21rc?

Re: cpufreq on Pentium M

[Dave Jones]

On Thu, Jun 12, 2003 at 10:21:27AM -0500, John Goerzen wrote:
 > On Thu, Jun 12, 2003 at 03:44:50PM +0100, Dave Jones wrote:
 > > On Thu, Jun 12, 2003 at 03:10:01PM +0100, Mark Watts wrote:
 > > 
 > >  > Thought this may help:
 > >  > - From a Dell D600 w/ 1.6Ghz Pentium M cpu...
 > > 
 > > Yep, the latest bunch of cache descriptors are missing from 2.4
 > > (They're in 2.5 already). I did send these to Marcelo, but they
 > > seem to have got lost when we entered -rc stage.
 > > I'll resend them for 2.4.22pre
 > 
 > Is it already too late for 2.4.21rc?

Yes. Exactly the same thing happened for .20rc, but with different
descriptors. Despite the change being so harmless its almost in the
'nothing can possibly go wrong' zone, Marcelo wanted to hold back.

		Dave

Re: cpufreq on Pentium M

[Martin List-Petersen]

[-- Attachment #1: Type: text/plain, Size: 322 bytes --]

On Thu, 2003-06-12 at 16:10, Mark Watts wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> 
> Thought this may help:
> - From a Dell D600 w/ 1.6Ghz Pentium M cpu...
> 

Then i don't need to resend this, i assume :)

Regards,
Martin List-Petersen
martin at list-petersen dot dk
--
Have at you!


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: cpufreq on Pentium M

[John Goerzen]

On Thu, Jun 12, 2003 at 07:18:03AM +0100, Dave Jones wrote:
> It should turn up in 2.5 sometime real soon, and at some point, maybe
> someone will backport it.

I do hope so.  I can be a risk taker but I've learned that running a
development kernel on a laptop does not make for a good travel experience. 
(When you've got no 'net connection and no backup device handy, getting the
latest patch for the IDE driver isn't fun <g>)

> Looks like missing cache descriptors. Grab x86info[1] and mail me
> the output of x86info -c

Here ya go:

heinrich:~# x86info -c
x86info v1.11.  Dave Jones 2001, 2002
Feedback to <davej@suse.de>.

Found 1 CPU
Family: 6 Model: 9 Stepping: 5 Type: 0
CPU Model: Unknown CPU Original OEM
unknown TLB/cache descriptor:
        0xb0
unknown TLB/cache descriptor:
        0xb3
Instruction TLB: 4MB pages, fully associative, 2 entries
unknown TLB/cache descriptor:
        0x87
unknown TLB/cache descriptor:
        0x30
Data TLB: 4MB pages, 4-way associative, 8 entries
unknown TLB/cache descriptor:
        0x2c

Re: cpufreq on Pentium M

[Dave Jones]

 > > Dave Jones <davej@codemonkey.org.uk> wrote:
 > > 
 > > Looks like missing cache descriptors. Grab x86info[1] and mail me
 > > the output of x86info -c
 > > 
 > > 		Dave
 > > 
 > > [1] http://www.codemonkey.org.uk/x86info/

For info folks, I've now had a dozen or so mails with the same cache
descriptors.  Please don't send me any more centrino based outputs,
I have plenty now, thanks 8-)

		Dave