Messages in this thread | | | Date | Tue, 15 Apr 2008 10:26:21 -0700 (PDT) | From | Linus Torvalds <> | Subject | Re: [PATCH] Replace completions with semaphores |
| |
On Tue, 15 Apr 2008, Andi Kleen wrote: > > For normal locks. But if you have N number of outstanding > events you need to wait for the semaphore is the right primitive.
Right. Except if you were to just use completions instead (which could be trivially extended to do that).
MUCH more trivial than this complex series.
(You may think that the "Replace completions with semaphores" patch is not very complicated, but it *is* - it depends very intimately on the big patch-series that basically turns semaphores into what completions are now!)
In other words, what makes me not like this is hat we first turn semaphores into the generic code (which is largely what completions were: just a special case of the generic semaphores!) and then turns completions into these things. That just doesn't make any sense to me!
Linus
| |