Re: [uml-devel] [patch 02/12] uml: cpu_relax fix

[Nick Piggin <nickpiggin@yahoo.com.au>]

Blaisorblade wrote:
> On Wednesday 23 March 2005 18:09, Bodo Stroesser wrote:
> 
>>blaisorblade@yahoo.it wrote:
>>
>>>Use rep_nop instead of barrier for cpu_relax, following $(SUBARCH)'s
>>>doing that (i.e. i386 and x86_64).
>>
>>IIRC, Jeff had the idea, to use sched_yield() for this (from a discussion
>>on #uml).
> 
> Hmm, makes sense, but this is to benchmark well... I remember from early 
> discussions on 2.6 scheduler that using sched_yield might decrease 
> performance (IIRC starve the calling application).
> 

Typically, for places where cpu_relax is used, sched_yield would be
a poor fit. So yes it could easily reduce performance.

> Also, that call should be put inside the idle loop, not for cpu_relax, which 
> is very different, since it is used (for instance) in kernel/spinlock.c for 
> spinlocks, and in such things. The "Pause" opcode is explicitly recommended 
> (by Intel manuals, I don't recall why) for things like spinlock loops, and 
> using yield there would be bad.
> 

The other thing is that sched_yield won't relax at all if you are the
only thing running, it will be a busy wait. So again, maybe not a great
fit for the idle loop either.