On Tue, Sep 19 2000, Andrea Arcangeli wrote:
> > 7[3] 8[2] 9[1] 10[0] 3[3] 4[2] 5[1] 6[0] 1[3] 2[2]
> p
> With point `p' I mean the request after last barrier in the queue.
Ah, I suspected we were talking past each other.
> Then when we try to insert 99 it goes here:
>
> 100[0] 102[3] 103[3] 104[3] 99[3]
> p
>
> So we have two low peaks in the not starving queue and we should move the p
> to the latest on the right.
Ok good, I've read Peter's patch now. Looks good, I've put it in my
tree as well and will do some testing.
> Also we should make different cases in function of what p->prev is
> (barrier/head/real_head/normalreq).
>
> I don't think it's worthwhile (even with the current algorithm where it's easy
> to account for p).
I suspect you are right, it's marginal.
-- * Jens Axboe <axboe@suse.de> * SuSE Labs - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 21:00:22 EST