powernow-k7, something broke in -test3?

powernow-k7, something broke in -test3?

[John Clemens]

powernow-k7's been running fine for me for a while... and it still seems
to work, but /proc/cpuinfo seems to suddenly be reporting weird results
since i upgraded from 2.5.75 to 2.6.0-test3 yesterday.  Any ideas? I'm
guessing this is cosmetic as my system still runs, but running the UAE
emulator seemed a bit sluggish today.. but that's probably my
imagination.

The numbers work out to something thinking i have ~77Mhz bus and not a
100Mhz one.

john.c -- whose starting to believe he's seeing things.

Here's a script output session:

Script started on Sun Aug 10 21:37:51 2003
root@diana:/usr/src# dmesg | grep powernow
powernow: AMD K7 CPU detected.
powernow: PowerNOW! Technology present. Can scale: frequency and voltage.
powernow: Found PSB header at c00f7ab0
powernow: Table version: 0x12
powernow: Flags: 0x0 (Mobile voltage regulator)
powernow: Settling Time: 100 microseconds.
powernow: Has 14 PST tables. (Only dumping ones relevant to this CPU).
powernow: PST:1 (@c00f7ad2)
powernow:  cpuid: 0x761 fsb: 100        maxFID: 0xb     startvid: 0xc
powernow:    FID: 0x10 (3.0x [300MHz])  VID: 0x11 (1.250V)
powernow:    FID: 0x4 (5.0x [500MHz])   VID: 0x11 (1.250V)
powernow:    FID: 0x6 (6.0x [600MHz])   VID: 0xd (1.350V)
powernow:    FID: 0x8 (7.0x [700MHz])   VID: 0xc (1.400V)
powernow:    FID: 0xb (8.5x [850MHz])   VID: 0xc (1.400V)
powernow: Minimum speed 300 MHz. Maximum speed 850 MHz.
root@diana:/usr/src# cat /proc/cpuinfo | grep MHz
cpu MHz         : 656.600
root@diana:/usr/src# cd /sys/devices/system/cpu/cpu0/cpufreq/
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_governor
userspace
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_setspeed
850000
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# echo "300000" >
scaling_setspeed
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_setspeed
300000
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat /proc/cpuinfo | grep
MHz
cpu MHz         : 231.741
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# echo "600000" >
scaling_setspeed
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat /proc/cpuinfo | grep
MHz
cpu MHz         : 463.482
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# echo "850000" >
scaling_setspeed
root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat /proc/cpuinfo |grep
MHz
cpu MHz         : 656.600
Script done on Sun Aug 10 21:41:14 2003

john.c

-- 
John Clemens          http://www.deater.net/john
john@deater.net     ICQ: 7175925, IM: PianoManO8
      "I Hate Quotes" -- Samuel L. Clemens

Re: powernow-k7, something broke in -test3?

[John Clemens]

Hi Dave,

The following patch corrects a minor think-o in your powernow-k7 changes
that went into 2.6.0-test3, and was causing the behavior below.  I don't
think you meant to cache the cfid value before you actually read it in :)
this was causing all speed notifications to think my previous multiplier
was fid_codes[0] (11.0x), and thus screws up loops_per_jiffy, etc..

john.c

diff -urN linux-vanilla/arch/i386/kernel/cpu/cpufreq/powernow-k7.c linux/arch/i386/kernel/cpu/cpufreq/powernow-k7.c
--- linux-vanilla/arch/i386/kernel/cpu/cpufreq/powernow-k7.c	2003-08-16 01:05:54.000000000 -0400
+++ linux/arch/i386/kernel/cpu/cpufreq/powernow-k7.c	2003-08-16 01:12:09.000000000 -0400
@@ -230,8 +230,8 @@

 	freqs.cpu = 0;

-	cfid = fidvidstatus.bits.CFID;
 	rdmsrl (MSR_K7_FID_VID_STATUS, fidvidstatus.val);
+	cfid = fidvidstatus.bits.CFID;
 	freqs.old = fsb * fid_codes[cfid] * 100;
 	freqs.new = powernow_table[index].frequency;




On Sun, 10 Aug 2003, John Clemens wrote:

>
> powernow-k7's been running fine for me for a while... and it still seems
> to work, but /proc/cpuinfo seems to suddenly be reporting weird results
> since i upgraded from 2.5.75 to 2.6.0-test3 yesterday.  Any ideas? I'm
> guessing this is cosmetic as my system still runs, but running the UAE
> emulator seemed a bit sluggish today.. but that's probably my
> imagination.
>
> The numbers work out to something thinking i have ~77Mhz bus and not a
> 100Mhz one.
>
> john.c -- whose starting to believe he's seeing things.
>
> Here's a script output session:
>
> Script started on Sun Aug 10 21:37:51 2003
> root@diana:/usr/src# dmesg | grep powernow
> powernow: AMD K7 CPU detected.
> powernow: PowerNOW! Technology present. Can scale: frequency and voltage.
> powernow: Found PSB header at c00f7ab0
> powernow: Table version: 0x12
> powernow: Flags: 0x0 (Mobile voltage regulator)
> powernow: Settling Time: 100 microseconds.
> powernow: Has 14 PST tables. (Only dumping ones relevant to this CPU).
> powernow: PST:1 (@c00f7ad2)
> powernow:  cpuid: 0x761 fsb: 100        maxFID: 0xb     startvid: 0xc
> powernow:    FID: 0x10 (3.0x [300MHz])  VID: 0x11 (1.250V)
> powernow:    FID: 0x4 (5.0x [500MHz])   VID: 0x11 (1.250V)
> powernow:    FID: 0x6 (6.0x [600MHz])   VID: 0xd (1.350V)
> powernow:    FID: 0x8 (7.0x [700MHz])   VID: 0xc (1.400V)
> powernow:    FID: 0xb (8.5x [850MHz])   VID: 0xc (1.400V)
> powernow: Minimum speed 300 MHz. Maximum speed 850 MHz.
> root@diana:/usr/src# cat /proc/cpuinfo | grep MHz
> cpu MHz         : 656.600
> root@diana:/usr/src# cd /sys/devices/system/cpu/cpu0/cpufreq/
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_governor
> userspace
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_setspeed
> 850000
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# echo "300000" >
> scaling_setspeed
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat scaling_setspeed
> 300000
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat /proc/cpuinfo | grep
> MHz
> cpu MHz         : 231.741
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# echo "600000" >
> scaling_setspeed
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat /proc/cpuinfo | grep
> MHz
> cpu MHz         : 463.482
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# echo "850000" >
> scaling_setspeed
> root@diana:/sys/devices/system/cpu/cpu0/cpufreq# cat /proc/cpuinfo |grep
> MHz
> cpu MHz         : 656.600
> Script done on Sun Aug 10 21:41:14 2003
>
> john.c
>
>

-- 
John Clemens          http://www.deater.net/john
john@deater.net     ICQ: 7175925, IM: PianoManO8
      "I Hate Quotes" -- Samuel L. Clemens

Re: powernow-k7, something broke in -test3?

[Dave Jones]

On Sat, Aug 16, 2003 at 01:24:58AM -0400, John Clemens wrote:
 > 
 > Hi Dave,
 > 
 > The following patch corrects a minor think-o in your powernow-k7 changes
 > that went into 2.6.0-test3, and was causing the behavior below.  I don't
 > think you meant to cache the cfid value before you actually read it in :)
 > this was causing all speed notifications to think my previous multiplier
 > was fid_codes[0] (11.0x), and thus screws up loops_per_jiffy, etc..

oh boy, how dumb am I.
Thanks, applied and pushed for 2.6-test4

		Dave

-- 
 Dave Jones     http://www.codemonkey.org.uk