[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] STAT bits - Ouch!

On Thu, 2010-12-09 at 22:31 +0100, Konrad Kokoszkiewicz wrote:
> >> I tend to disagree, I bet the gain wouldn't be even measurable.
> >
> > And I disagree with you on this. The fact you are saving CPU cycles
> > means improved performance. Period. Whether it's measurable or not
> > depends on what you are doing.
> Saving CPU cycles is not a good thing per se, it is only good when it 
> saves time. When you sacrifice a measurable time to make an unmeasurable 
> execution gain, you loose time, not gain it.

And your point is?

> In this case, any minute you waste on this is worth 19 millions 200 
> thousands times the gain you expect to make (on standard 16 MHz Falcon).
> And you still didn't provide any more example than the two programs which 
> apparently break the POSIX standard by relying on values they should not 
> rely on.

I didn't say this. POSIX doesn't define these values. 

> If anyone asks me, there is no evidence MiNT's stat() needs change. This 
> is the two programs which need fixing.

So you didn't really read my initial email properly.

Unless someone stumps up the time to change git, then that application
will never work. Python was an easy fix. There may be other applications
out there too, who knows. But until you actually bump into the problem
and go hunting it can be a needle in a haystack.

If we change (and I'm not saying need to), then application level fixing
and therefore manpower on fixing other applications would be a zero
effort for making the changes to the kernel.

And obviously Helmut has noticed your overly negative attitude towards
the thread. And as I've mentioned before, I'll leave it for now and
gather any more applications I bump into.

I'm not spending any more time on this thread.