Re: [PATCH 5/6] ARM: OMAP2: drop redundant pending write check for gptimer

[Tony Lindgren <tony@atomide.com>]

* Russell King - ARM Linux <linux@arm.linux.org.uk> [081207 09:06]:
> On Sun, Dec 07, 2008 at 08:55:54AM -0600, Woodruff, Richard wrote:
> > Yes, NOHZ is _poor_ today in respect to needless reprogramming. Code can
> > be improved. I have sent Thomas a patch for the same which is in the MM
> > tree for a while now.
> 
> When I discussed this problem with Thomas, his reaction was rather
> negative due to there being conflicting requirements between this and
> ia64.  And quite frankly I think that the OMAP behaviour is utter
> madness - taking as long as it does to update the timer registers
> means that no matter what you do, you _will_ be paying a penalty
> for that hardware.

Well still the timer code should be optimized for slower systems.
The timer should not be reprogrammed during every device interrupt
if the change is less than a jiffy. And going through the
timer lists is not always needed, so Richard's patch tries to
optimize that [1].

> Let's hope that later OMAP designs sanitise the timer hardware so
> it doesn't take stupid amounts of time to write to the registers,
> posted mode or not.

Yeah. Richard has some profiling data showing you how much time is
spent doing that, maybe post that again for reference? It's very
interesting data.

Tony


[1] http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.28-rc2/2.6.28-rc2-mm1/broken-out/tick-schedc-suppress-needless-timer-reprogramming.patch