Date:	Tue, 12 Dec 2006 08:47:46 -0600
From:	Maynard Johnson <maynardj@...ibm.com>
To:	Luke Browning <LukeBrowning@...ibm.com>
CC:	maynardj@...ux.vnet.ibm.com,
	Arnd Bergmann <arnd.bergmann@...ibm.com>,
	cbe-oss-dev@...abs.org, linux-kernel@...r.kernel.org,
	linuxppc-dev@...abs.org, oprofile-list@...ts.sourceforge.net
Subject: Re: [Cbe-oss-dev] [PATCH]Add notification for active Cell SPU tasks

Luke Browning wrote:
> maynardj@...ux.vnet.ibm.com wrote on 08/12/2006 01:04:30 PM:
> 
>  > Arnd Bergmann wrote:
>  >
>  > >On Wednesday 06 December 2006 23:04, Maynard Johnson wrote:
>  > >  
>  > >No code should ever need to look at other SPUs when performing an
>  > >operation on a given SPU, so we don't need to hold a global lock
>  > >during normal operation.
>  > >
>  > >We have two cases that need to be handled:
>  > >
>  > >- on each context unload and load (both for a full switch operation),
>  > >  call to the profiling code with a pointer to the current context
>  > >  and spu (context is NULL when unloading).
>  > >
>  > >  If the new context is not know yet, scan its overlay table (expensive)
>  > >  and store information about it in an oprofile private object. 
> Otherwise
>  > >  just point to the currently active object, this should be really 
> cheap.
>  > >
>  > >- When enabling oprofile initially, scan all contexts that are currently
>  > >  running on one of the SPUs. This is also expensive, but should happen
>  > >  before the measurement starts so it does not impact the resulting 
> data.
>  > >
> 
> Agreed.  
> 
> <snip>
> 
>  > >>I'm not exactly sure what you're saying here.  Are you suggesting 
> that a
>  > >>user may only be interested in acitve SPU notification and, therefore,
>  > >>shouldn't have to be depenent on the "standard" notification
>  > >>registration succeeding?  There may be a case for adding a new
>  > >>registration function, I suppose; although, I'm not aware of any other
>  > >>users of the SPUFS notification mechanism besides OProfile and PDT, 
> and
>  > >>we need notification of both active and future SPU tasks.  But I would
>  > >>not object to a new function.
>  > >>
>  > >>    
>  > >>
>  > >I think what Luke was trying to get to is that notify_spus_active() 
> should
>  > >not call blocking_notifier_call_chain(), since it will notify other 
> users
>  > >as well as the newly registered one. Instead, it can simply call the
>  > >notifier function directly.
>  > >  
>  > >
>  > Ah, yes.  Thanks to both of you for pointing that out.  I'll fix that
>  > and re-post.
>  >
>  > -Maynard
>  >
> 
> I actually was hoping to take this one step further.  If the interface to
> the context switch handler is something like:
> 
> switch_handler(int spu_id, from_ctx, to_ctx)
The function prototype for the switch handler is set in concrete by the 
notification framework.  The parameters are: struct notifier_block *, 
unsigned long, void *.
> 
> The kernel extension can maintain an internal spu table of its own where it
> marks the named spuid as active or not.  You don't need to have a bunch of
> individual calls.  Internally, you can keep track of it yourself.
I think this would be nice to have, and I will look into it as I have 
time.  However, for the existing usage of the SPU switch notification, I 
don't think it's too critical, since most users are not going to be 
trying to do profiling or debugging with multiple SPU apps running 
simultaneously.
> 
> Luke
> 


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists