linuxppc-dev.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
* Re: get_cycles()
  2008-04-26  7:54 get_cycles() Kevin Diggs
@ 2008-04-26  7:11 ` David Miller
  2008-04-27  2:39   ` get_cycles() Kevin Diggs
  0 siblings, 1 reply; 6+ messages in thread
From: David Miller @ 2008-04-26  7:11 UTC (permalink / raw)
  To: kevdig; +Cc: linuxppc-dev

From: Kevin Diggs <kevdig@hypersurf.com>
Date: Sat, 26 Apr 2008 00:54:11 -0700

> 	Anyone know how to turn get_cycles() into an actual time in a module? 
> ppc_tb_freq does not seem to be exported?

You should really be using ktime_t and associated interfaces.

^ permalink raw reply	[flat|nested] 6+ messages in thread

* get_cycles()
@ 2008-04-26  7:54 Kevin Diggs
  2008-04-26  7:11 ` get_cycles() David Miller
  0 siblings, 1 reply; 6+ messages in thread
From: Kevin Diggs @ 2008-04-26  7:54 UTC (permalink / raw)
  To: linuxppc-dev

H,

	Anyone know how to turn get_cycles() into an actual time in a module? 
ppc_tb_freq does not seem to be exported?

kevin

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: get_cycles()
  2008-04-26  7:11 ` get_cycles() David Miller
@ 2008-04-27  2:39   ` Kevin Diggs
  2008-04-27  3:21     ` get_cycles() David Miller
  0 siblings, 1 reply; 6+ messages in thread
From: Kevin Diggs @ 2008-04-27  2:39 UTC (permalink / raw)
  To: linuxppc-dev

David Miller wrote:
> From: Kevin Diggs <kevdig@hypersurf.com>
> Date: Sat, 26 Apr 2008 00:54:11 -0700
> 
> 
>>	Anyone know how to turn get_cycles() into an actual time in a module? 
>>ppc_tb_freq does not seem to be exported?
> 
> 
> You should really be using ktime_t and associated interfaces.
> 
This looks pretty cool ... but I don't want to create a dependency on
hrtimer.

So ... How expensive would it be to export ppc_tb_freq? Or add a
get_cycles_tb() function?

kevin

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: get_cycles()
  2008-04-27  2:39   ` get_cycles() Kevin Diggs
@ 2008-04-27  3:21     ` David Miller
  2008-04-27 10:35       ` get_cycles() Kevin Diggs
  0 siblings, 1 reply; 6+ messages in thread
From: David Miller @ 2008-04-27  3:21 UTC (permalink / raw)
  To: kevdig; +Cc: linuxppc-dev

From: Kevin Diggs <kevdig@hypersurf.com>
Date: Sat, 26 Apr 2008 19:39:07 -0700

> This looks pretty cool ... but I don't want to create a dependency on
> hrtimer.

It doesn't create such a dependency.

We use it unconditionally in the generic networking.

Please don't use platform specific interfaces if you don't have to.
You're be insulated from so many things.

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: get_cycles()
  2008-04-27  3:21     ` get_cycles() David Miller
@ 2008-04-27 10:35       ` Kevin Diggs
  2008-04-28  4:38         ` get_cycles() Tony Breeds
  0 siblings, 1 reply; 6+ messages in thread
From: Kevin Diggs @ 2008-04-27 10:35 UTC (permalink / raw)
  To: linuxppc-dev

David Miller wrote:
> From: Kevin Diggs <kevdig@hypersurf.com>
> Date: Sat, 26 Apr 2008 19:39:07 -0700
> 
> 
>>This looks pretty cool ... but I don't want to create a dependency on
>>hrtimer.
> 
> 
> It doesn't create such a dependency.
> 
> We use it unconditionally in the generic networking.
> 
> Please don't use platform specific interfaces if you don't have to.
> You're be insulated from so many things.
> 

I'm working on a cpufreq driver for the 750GX so I don't think I have to 
worry about being to platform specific.

Would a compile time configuration be a good idea (hrtimer or 
get_cycles() assisted timing)?

In the 2.4 code I just used a timer 2 ticks in the future to be certain 
I did not go under the 100 us PLL lock delay. I was trying to see if I 
could cut the latency down.

What about using OF? Isn't there a timebase property for the cpus?

Thoughts?

kevin

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: get_cycles()
  2008-04-27 10:35       ` get_cycles() Kevin Diggs
@ 2008-04-28  4:38         ` Tony Breeds
  0 siblings, 0 replies; 6+ messages in thread
From: Tony Breeds @ 2008-04-28  4:38 UTC (permalink / raw)
  To: Kevin Diggs; +Cc: linuxppc-dev

On Sun, Apr 27, 2008 at 03:35:54AM -0700, Kevin Diggs wrote:

> I'm working on a cpufreq driver for the 750GX so I don't think I have to 
> worry about being to platform specific.
> 
> Would a compile time configuration be a good idea (hrtimer or 
> get_cycles() assisted timing)?

Save yourself the pain, and use the hrtimers infrastructure.
 
> In the 2.4 code I just used a timer 2 ticks in the future to be certain 
> I did not go under the 100 us PLL lock delay. I was trying to see if I 
> could cut the latency down.
> 
> What about using OF? Isn't there a timebase property for the cpus?

Sure, If you really don't want to use the infrastructure that's there
feel free to duplicate portions of arch/powerpc/kernel/time.c and
kernel/time/* into your cpufreq driver.

Yours Tony

  linux.conf.au    http://www.marchsouth.org/
  Jan 19 - 24 2009 The Australian Linux Technical Conference!

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2008-04-28  4:38 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-04-26  7:54 get_cycles() Kevin Diggs
2008-04-26  7:11 ` get_cycles() David Miller
2008-04-27  2:39   ` get_cycles() Kevin Diggs
2008-04-27  3:21     ` get_cycles() David Miller
2008-04-27 10:35       ` get_cycles() Kevin Diggs
2008-04-28  4:38         ` get_cycles() Tony Breeds

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).