[speedstep] testers wanted (Jochen Gensch)

[Eric Kjeldergaard]

> Hi,
> 
> On Thu, Sep 16, 2004 at 10:39:02AM -0500, kjelderg at sponge.dyndns.org wrote:
> > >  > I just finished a speedstep driver for p3m and p4m but it's a little 
bit
> > >  > ugly, and worst it would work only with ich chipsets.  Even worst, I
> > >  > don't have the hardware so I can't test it.
> > >
> > I'd also be happy to test if you send exactly what you want 
tested.  Thinkpad 
> > R40 with p4m here.  
> 
> Yes, I was not exactly enough explicit, sorry..
> 
> Something like that, 
> 
> sysctl dev.ichist.0.speedstep
> Should give 0 (if high frequency) or 1 (if low frequency),
> then
> openssl speed sha1
> sysctl dev.ichist.0.speedstep=(0 or 1, in order to change frequency)
> then redo a
> openssl speed sha1
> 
> to see if something change.
> 
> Thanks,
> 
> -- 
> Bruno Ducrot
> 
> --  Which is worse:  ignorance or apathy?
> --  Don't know.  Don't care.
makes and installs fine.  Detechts ICH4m bridge perfectly.

bash-3.00# sysctl -a | grep ichist
dev.ichist.0.%desc: Intel ICH4m LPC bridge
dev.ichist.0.%driver: ichist
dev.ichist.0.%parent: nexus0
dev.ichist.0.speedstep: 1

And now for a few quick benchmarks to show it works:
dev.ichist.0.speedstep: 0 -> 0
bash-3.00# openssl speed sha1
To get the most accurate results, try to run this
program when this computer is idle.
Doing sha1 for 3s on 16 size blocks: 1433878 sha1's in 2.90s
Doing sha1 for 3s on 64 size blocks: 984912 sha1's in 2.86s
Doing sha1 for 3s on 256 size blocks: 509682 sha1's in 2.87s
Doing sha1 for 3s on 1024 size blocks: 170379 sha1's in 2.86s
Doing sha1 for 3s on 8192 size blocks: 23581 sha1's in 2.89s
OpenSSL 0.9.7d 17 Mar 2004
built on: Tue Sep 21 01:07:22 CDT 2004
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) 
blowfish(idx)
compiler: cc
available timing options: USE_TOD HZ=128 [sysconf value]
timing function used: getrusage
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
sha1              7898.60k    22040.86k    45465.74k    60943.57k    66811.31k
bash-3.00# sysctl dev.ichist.0.speedstep=1
dev.ichist.0.speedstep: 0 -> 1
bash-3.00# openssl speed sha1
To get the most accurate results, try to run this
program when this computer is idle.
Doing sha1 for 3s on 16 size blocks: 768526 sha1's in 2.85s
Doing sha1 for 3s on 64 size blocks: 537934 sha1's in 2.87s
Doing sha1 for 3s on 256 size blocks: 282610 sha1's in 2.86s
Doing sha1 for 3s on 1024 size blocks: 98091 sha1's in 2.87s
Doing sha1 for 3s on 8192 size blocks: 13662 sha1's in 2.84s
OpenSSL 0.9.7d 17 Mar 2004
built on: Tue Sep 21 01:07:22 CDT 2004
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) 
blowfish(idx)
compiler: cc
available timing options: USE_TOD HZ=128 [sysconf value]
timing function used: getrusage
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
sha1              4308.26k    11980.80k    25275.55k    34956.50k    39435.07k
bash-3.00# sysctl dev.ichist.0.speedstep=0
dev.ichist.0.speedstep: 1 -> 0
bash-3.00# openssl speed sha1
To get the most accurate results, try to run this
program when this computer is idle.
Doing sha1 for 3s on 16 size blocks: 1431234 sha1's in 2.90s
Doing sha1 for 3s on 64 size blocks: 1000531 sha1's in 2.91s
Doing sha1 for 3s on 256 size blocks: 528025 sha1's in 2.91s
Doing sha1 for 3s on 1024 size blocks: 181941 sha1's in 2.89s
Doing sha1 for 3s on 8192 size blocks: 25287 sha1's in 2.91s
OpenSSL 0.9.7d 17 Mar 2004
built on: Tue Sep 21 01:07:22 CDT 2004
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) 
blowfish(idx)
compiler: cc
available timing options: USE_TOD HZ=128 [sysconf value]
timing function used: getrusage
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
sha1              7890.36k    22019.45k    46426.83k    64513.28k    71261.79k
bash-3.00# sysctl dev.ichist.0.speedstep=1
dev.ichist.0.speedstep: 0 -> 1
bash-3.00# openssl speed sha1
To get the most accurate results, try to run this
program when this computer is idle.
Doing sha1 for 3s on 16 size blocks: 751338 sha1's in 2.78s
Doing sha1 for 3s on 64 size blocks: 471788 sha1's in 2.52s
Doing sha1 for 3s on 256 size blocks: 283272 sha1's in 2.87s
Doing sha1 for 3s on 1024 size blocks: 97830 sha1's in 2.87s
Doing sha1 for 3s on 8192 size blocks: 13755 sha1's in 2.84s
OpenSSL 0.9.7d 17 Mar 2004
built on: Tue Sep 21 01:07:22 CDT 2004
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) 
blowfish(idx)
compiler: cc
available timing options: USE_TOD HZ=128 [sysconf value]
timing function used: getrusage
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
sha1              4323.55k    12000.71k    25240.58k    34927.47k    39729.25k
bash-3.00#

Looks very promising.  Is the eventual plan (as Linux does) to run a daemon 
that modifies this setting, or as I had rather though was happening, to move 
acpi downclocking into a kernel module?  Either way, is there any word on the 
implementation this particular feature?