Re: Thoughts and Opinions or something like that

[Keith Thompson <kst@cts.com>]

jmv16@cornell.edu (Jeffrey M. Vinocur) writes:
[...]
> Well, the goal isn't to lessen branch prediction -- the silicon
> is there either way -- but rather to reduce mispredicted
> branches.  The most basic approach, I guess, is assume all
> backwards branches will be taken -- this gets loops.  One
> not-very-complicated approach is to cache whether a branch was
> taken last time through and predict the same.  This approach,
> however, does exactly what you suggest; it gets loop iterations
> 2,3,4,...,n-1 all correct.

Note that if you have multiple threads/processes/coroutines/whatever,
the threads can interfere with each other's cached branch information.

In practice, I suspect this doesn't make much difference.  You could
cache the branch information separately in each thread, but I suspect
this wouldn't be worth the effort.  It might even be harmful if it
prevents one thread from "learning" from the experience of another.

-- 
Keith Thompson (The_Other_Keith) kst@cts.com  <http://www.ghoti.net/~kst>
San Diego Supercomputer Center           <*>  <http://www.sdsc.edu/~kst>
Cxiuj via bazo apartenas ni.