Re: [ANNOUNCE] ktimers subsystem

[Thomas Gleixner <tglx@linutronix.de>]

On Mon, 2005-09-19 at 16:04 -0700, Christoph Lameter wrote:
> On Tue, 20 Sep 2005, Thomas Gleixner wrote:
> 
> > Hmm. I don't understand the argument line completely. 
> > 
> > 1. The kernel has to provide ugly mechanisms because a lot of
> > applications implementations are doing the Wrong Thing ?
> 
> Lets skip the "wrong thing"... Or are you saying that glibc and all the 
> apps are all wrong? 
> 
> Applications call gettimeofday for a variety of reasons. One is because it 
> is widely available over different platformsn and application want to 
> schedule things, need timestamps etc etc.

Accepted. But I still doubt that the number of calls to gettimeofday is
in anyway justified. The question I'm asking if it is really worth a
long and epic discussion about a single add instruction ?


> > > Many platforms can execute gettimeofday 
> > > without having to enter the kernel.
> > 
> > Which ones ? How is this achieved with respect to all the time adjust,
> > correction... code ?
> 
> IA64 f.e. has a special instruction that allows access to kernel user 
> space without having to do a context switch.

Ok, was not aware of that and John kindly clarified this already. 

tglx