Re: Make condition variables take absolute wakeup time

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Wed, May 07, 2008 at 06:42:51PM +0200, Dieter Baron wrote:
> In article <200805071605.MAA14765%Sparkle.Rodents.Montreal.QC.CA@localhost> 
> der wrote:
> : > These days, does hz >> 2 ever generate better code than hz / 4?
> 
> : <kernel.h> says hz is signed, so, if shifts are faster than divides, I
> : would expect >>2 to be faster than /4.  If the hardware doesn't have a
> : signed-integer divide but does have shifts (eg, least common
> : denominator sparc32), >>2 should be smaller than /4 too.
> 
>   Actually, I would expect modern compilers to know that and convert
> /4 into >>2 internally, when advantageous.  IIUC, we are talking about
> constants here.

The shift conversion is only for unsigned arithmetic trival.

Joerg