On Tue, 26 Feb 2013, Magnus Damm wrote:

On Tue, Feb 26, 2013 at 2:16 AM, Guennadi Liakhovetski
[off-list ref] wrote:

quoted

In v2 of this patch-set I dropped patches 1 and 3, because #1 is a bug-fix
and can go in separately, #3 is more of a debugging help, than a real
patch for mainlining. Of the remaining 2 patches the main difference to v1
is the omission of the PLL0 manipulation. More is explained in respective
patches.

Guennadi Liakhovetski (2):
  ARM: shmobile: sh73a0: add support for adjusting CPU frequency
  ARM: shmobile: kzm9g-reference: add CPUFreq support

Hi Guennadi,

Thanks for your work on this. In general I am quite happy about this series.

One rather significant question is however this: Would it be possible
for you to check if the TWD is operating as expected when you scale
the CPU Frequency? I suspect it is not, so please cook up a v3 of this
series that also includes updates for timers.

At last I managed to get twd running with a recent (of 27.02) all+next 
snapshot on kzm9g with DT (reference) and actually I'm not really sure the 
problem is there. At least definitely not as bad as I saw it last time. To 
remind, the problem I observed was, that at lower CPU speeds a simple 
shell script, consisting of 20 20-second sleep calls was gaining multiple 
seconds, was taking more like 405 seconds, than 400. Now I only see it 
going to 401 seconds. This is still not good, but I need a more precise 
comparison to actually decide, whether this problem only appears at lower 
CPU speeds or it is always there. I'll look further into it next week.

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/