Re: est.c patch for Pentium-M 1.86GHz

[Victor Balada Diaz <victor@xxxxxxxxx>]

On Sun, Dec 10, 2006 at 12:33:26AM +0900, YONETANI Tomokazu wrote:
> On Fri, Dec 08, 2006 at 05:08:03PM +0100, Victor Balada Diaz wrote:
> > On Wed, Dec 06, 2006 at 01:25:10AM +0100, Victor Balada Diaz wrote:
> > > On Wed, Dec 06, 2006 at 08:34:40AM +0900, YONETANI Tomokazu wrote:
> > > > By the way if you saw a message like this with unpatched est driver,
> > > > please let me know:
> > > > ... - unknown CPU or operating point(cpu_id:XXXXX, msr:YYYY).
> > > 
> > > Enhanced SpeedStep (1308 mV)  - unknown CPU or operating point(cpu_id:0x6d8, msr:0x6120f2606000f26).
> > > module_register_init: MOD_LOAD (est, d56040e8, 0) error 45
> > > 
> > > It's a Pentium M 760 @ 2GHZ.
> > 
> > With the attached patch it does work. I've got it from
> > FreeBSD's src/sys/i386/cpufreq/est.c. The model i pick up was 
> > static freq_info PM_755C_90[] = {
> > 	/* 90 nm 2.00GHz Pentium M, VID #C */
> 
> Are you sure?  Decoding the msr value above, your frequency/voltage
> range is (800MHz, 988mV) - (2000MHz, 1308mV), so I guess multiple
> entries(600MHz and 800MHz, for example) end up in the same CPU
> performance.  NetBSD people seem to be using `openssl speed' to
> measure the performance for each setting.

No, i wasn't sure. Seems you're right and some frequencies are the same:

600 == 800 == 1000

patata# sysctl machdep.est.frequency.target=600
machdep.est.frequency.target: 1333 -> 533
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     19269.92k    20835.95k    21156.42k    21298.59k    21296.05k


patata# sysctl machdep.est.frequency.target=800
machdep.est.frequency.target: 533 -> 800
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     19241.44k    20783.55k    21157.00k    21187.33k    21294.18k


patata# sysctl machdep.est.frequency.target=1000
machdep.est.frequency.target: 800 -> 933
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     19207.97k    20744.01k    21079.12k    21242.23k    21346.75k


patata# sysctl machdep.est.frequency.target=1200
machdep.est.frequency.target: 933 -> 1200
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     28905.60k    31264.93k    31755.43k    31957.56k    31952.19k


patata# sysctl machdep.est.frequency.target=1400
machdep.est.frequency.target: 1200 -> 1333
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     32053.43k    34439.21k    35263.81k    35338.98k    35465.83k


patata# sysctl machdep.est.frequency.target=1600
machdep.est.frequency.target: 1333 -> 1600
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     38342.31k    41485.92k    42213.80k    42522.47k    42497.27k


patata#  sysctl machdep.est.frequency.target=1800
machdep.est.frequency.target: 1600 -> 1733
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     40962.07k    44575.12k    45788.80k    45948.40k    46160.46k


patata#sysctl machdep.est.frequency.target=2000
machdep.est.frequency.target: 1733 -> 2000
patata# openssl speed blowfish
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     48181.66k    52122.45k    52936.15k    53145.65k    53413.34k


-- 
La prueba más fehaciente de que existe vida inteligente en otros
planetas, es que no han intentado contactar con nosotros.