← Back to team overview

ac100 team mailing list archive

Re: Stability Under Load


On 08/20/2011 03:13 PM, Julian Andres Klode wrote:
On Sat, Aug 20, 2011 at 10:12:58AM +0100, Gordan Bobic wrote:
On 08/20/2011 09:28 AM, Marc Dietrich wrote:
Am Freitag 19 August 2011, 20:51:29 schrieb Gordan Bobic:
   Just looking now. Are you referring to the REGULATOR_INIT macro and
   specifically the line that defines a struct with:

   REGULATOR_INIT(sm1, 725, 1200, true)?

   Are you saying that the default value as provided by Toshiba was

   REGULATOR_INIT(sm1, 725, 1000, true)?

correct. Please understand that these values are just best guesses on
what the Toshiba code does. In fact, kernel 2.6.38 has nothing to do
with the original code base because it's a total rewrite. You may check
tegra/odm_kit/adaptations/pmu/tps6586x/nvodm_pmu_tps6586x.c for

I don't seem to have the arch/arm/mach-tegra/odm_kit directory in the
kernel tree I pulled from your git. Where can I get this?

>from the original source (check .29 or .32 kernel on gitorious.org).

Btw, I checked again and it seems that I only changed SM1. So it would be great
if you can do some stress test with both voltages. I'm also not sure what
voltages are really used in the original kernel. The code is sometimes a bit
confusing. But you can check by running .29 and look in /sys/devices/regulator
(I guess).

OK, I'll take a look at that. My new AC100 passed 9 hours of pbzip2
testing without any errors, so the evidence so far is that putting
SM1 back to 1000mV did stabilize it. Testing with the old one (the
one that was very unstable) now - will report back tonight.
I just tried to build a kernel with one restricted to 1000, but quickly
got a build failure. It might be caused by having tried to build a
kernel using a 1200 previously and the device still being a bit messed
up from that, though.

I should give it some rest and try again in a few hours.

Which kernel are you using?

I just did the change to board-paz00-power.c reducing SM1 voltage from 1200mV to 1000mV, and now I have been running the pbzip2 test on a loop at low priority and recompiling glibc at normal priority, and I've not been able to shake either of my AC100s loose. Since I couldn't get anywhere near this sort of load before without something erroring out, I'd say that this made quite a substantial difference to stability.

Marc, could you please change this to 1000mV in git? Running at 1200mV really seems to upset stability.


Follow ups