[Openmcl-devel] Quick HW question...
dlw at itasoftware.com
Wed Nov 17 19:29:59 UTC 2010
Gary Byers wrote:
> Exactly what needs to be done to help the processor predict whether a
> is taken tends to be architecture- and implementation-dependent.
> (It's often
> the case that simple rules - like "backward conditional branches are
> to be taken and forward conditional branches are predicted to not be
> taken" -
> go a long way.)
Do current architecures do dynamic branch predication, in which
they keep a bit at runtime to keep track of which way the
branch went last time? I know that such things have,
at the very least, been suggested, although I had no idea
to what extent they have been put into use, and whether
measurements show that they are enough better than
static predication to justify their cost.
> Suppose that there were semi-magic macros (I -think- that they could
> be macros)
> called something like EXPECTED-TRUE and EXPECTED-FALSE
Yeah, I always imgined "jump-usually" "if-usually" too.
More information about the Openmcl-devel