Messages in this thread | | | Date | Mon, 4 Jun 2007 16:28:55 +0200 | From | Eric Dumazet <> | Subject | Re: [patch 1/2] ufd v1 - unsequential O(1) fdmap core |
| |
On Mon, 4 Jun 2007 06:35:16 -0700 (PDT) Davide Libenzi <davidel@xmailserver.org> wrote:
> On Mon, 4 Jun 2007, Eric Dumazet wrote: > > > You add conditional branches on very hot spots. > > Keep BS for the ones you argue usually, and that are not able to reply. > You *still* two bitmaps, because allocation spaces are far apart. So the > "if" will still be there.
I actually read your patches and spent time to see the pros and cons.
If you dont need reviewers, please dont post your patches on lkml.
If I am not mistaken, you added a test in fget()/fget_light(), which is a known hot point for said huge processes.
fget() dont need to access the bitmap at all. Using fd_slots means less (50%) file pointers per cache line.
On my machines, there is a ratio of 100/1 in cpu time for fget(),fget_light() against get_unused_fd().
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |