From owner-freebsd-alpha Sat Jan 27 19:48:50 2001 Delivered-To: freebsd-alpha@freebsd.org Received: from duke.cs.duke.edu (duke.cs.duke.edu [152.3.140.1]) by hub.freebsd.org (Postfix) with ESMTP id 0965737B400; Sat, 27 Jan 2001 19:48:24 -0800 (PST) Received: from grasshopper.cs.duke.edu (grasshopper.cs.duke.edu [152.3.145.30]) by duke.cs.duke.edu (8.9.3/8.9.3) with ESMTP id WAA27625; Sat, 27 Jan 2001 22:48:19 -0500 (EST) Received: (from gallatin@localhost) by grasshopper.cs.duke.edu (8.11.1/8.9.1) id f0S3mJ530049; Sat, 27 Jan 2001 22:48:19 -0500 (EST) (envelope-from gallatin@cs.duke.edu) From: Andrew Gallatin MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Date: Sat, 27 Jan 2001 22:48:19 -0500 (EST) To: John Baldwin Cc: Andrew Gallatin , freebsd-alpha@FreeBSD.org, Bernd Walter Subject: Re: -current failing on PC164 In-Reply-To: References: <14963.12976.121070.21427@grasshopper.cs.duke.edu> X-Mailer: VM 6.43 under 20.4 "Emerald" XEmacs Lucid Message-ID: <14963.38369.912484.545380@grasshopper.cs.duke.edu> Sender: owner-freebsd-alpha@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org John Baldwin writes: > > On 27-Jan-01 Andrew Gallatin wrote: > > > > FWIW, the perf. decrease you noticed is probably directly related to > > the mutex stuff no longer being inlined -- function calls are pretty > > expensive on alpha. I saw a huge perf. increase when we inlined spl's > > last year. > > Right now the mutex API is being somewhat overhauled. When it is done, we may > move back to using inlining. If it is a big performance boost on the alpha, > then we will definitely inline on at least the alpha. For x86, function calls > are (relatively) cheap, so we weren't as worried about it. I am not as aware > however when it comes to alpha specifics. Please feel free to knock me upside > the head when we do something stupid. Will do. I realize that perf. is not the goal right now, o/w I'd have been screaming my head off long ago ;) Once perf. does become an issue, we'll have to get you using Iprobe. Its one hell of a lot better than gprof (sampling with 5% overhead, rather than doctoring each function call with 50% overhead..). Cheers, Drew To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message