> "Theodore Y. Ts'o" <tytso@MIT.EDU> writes:
> > (1 << OUT_FLAG_NUMBER) : 0") both involve an assembly language
> > conditional jump instruction, which is expensive, because it involves
> > pipeline stalls, etc.
>
> Has it been measured ? I know that branch prediction is not 100%, but
> I'm just trying to make sure we're not trying to solve a non-problem (if the
> prediction works well, then the jump costs 0 (zero) cycles).
If you have (working) branch prediction the direct cost will be
negligible, if any. But the indirect cost is that you use a slot
in the branch table which means you might well take the full cost
somewhere else later...
Mike
-- .----------------------------------------------------------------------. | Mike Jagdis | Internet: mailto:mike@roan.co.uk | | Roan Technology Ltd. | | | 54A Peach Street, Wokingham | Telephone: +44 118 989 0403 | | RG40 1XG, ENGLAND | Fax: +44 118 989 1195 | `----------------------------------------------------------------------'