From owner-freebsd-current Sun Jun 27 11:10: 2 1999 Delivered-To: freebsd-current@freebsd.org Received: from apollo.backplane.com (apollo.backplane.com [209.157.86.2]) by hub.freebsd.org (Postfix) with ESMTP id 1631615068 for ; Sun, 27 Jun 1999 11:09:59 -0700 (PDT) (envelope-from dillon@apollo.backplane.com) Received: (from dillon@localhost) by apollo.backplane.com (8.9.3/8.9.1) id LAA14643; Sun, 27 Jun 1999 11:09:56 -0700 (PDT) (envelope-from dillon) Date: Sun, 27 Jun 1999 11:09:56 -0700 (PDT) From: Matthew Dillon Message-Id: <199906271809.LAA14643@apollo.backplane.com> To: Doug Rabson Cc: Peter Wemm , current@freebsd.org, mckusick@mckusick.com Subject: Re: BUF_LOCK() related panic.. References: Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG :In the long term, we probably need an spl-aware simplelock or maybe the :cunning no-cost interrupt thread scheme which BSDi are using. : :-- :Doug Rabson Mail: dfr@nlsystems.com :Nonlinear Systems Ltd. Phone: +44 181 442 9037 I really like the idea of a cunning no-cost interrupt thread scheme, I didn't realize that BSDi had moved to it! Doing this would allow us to remove easily half the spl management code and would allow us to fix simplelocks - to implement them the way they ought to be implemented. -- On a related topic, I've done some syscall overhead measurements w/ SMP verses non-SMP systems that people may be interested in. The SMP/patched version includes a patch that Alan is going to commit soon ( a relatively simple removal of unnecessary locks surrounding an already-atomic assembly operation ). The cpu's are P-III/450's. no SMP Null syscall: 3532 nanoseconds SMP Null syscall: 6750 nanoseconds SMP/patched Null syscall: 6533 nanoseconds In comparison, I believe the Linux syscall overhead is on the order of 1 uS. Not that I am concerned about a few measily microseconds, but this does show that we have a ways to go in our SMP work. There is a lot of potential for improvement. -Matt Matthew Dillon To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message