From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dmitry Torokhov Subject: Re: [RFC] Should we create a raw input interface for IR's ? Date: Thu, 26 Nov 2009 15:28:09 -0800 Message-ID: <20091126232809.GE6936@core.coreip.homeip.net> References: <829197380911251020y6f330f15mba32920ac63e97d3@mail.gmail.com> <20091126055302.GI23244@core.coreip.homeip.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Krzysztof Halasa Cc: Jarod Wilson , Devin Heitmueller , Christoph Bartelmus , awalls@radix.net, j@jannau.net, jarod@redhat.com, linux-input@vger.kernel.org, linux-kernel@vger.kernel.org, linux-media@vger.kernel.org, mchehab@redhat.com, superm1@ubuntu.com List-Id: linux-input@vger.kernel.org On Thu, Nov 26, 2009 at 07:40:01PM +0100, Krzysztof Halasa wrote: > Dmitry Torokhov writes: > > > Why would sysfs write be slower than ioctl? > > Sysfs is generally one-value, one-file. open, read/write, close. > ioctl() OTOH does everything (e.g. a whole key table) in one syscall. There are binary sysfs attributes. For ioctl you also need to open and close the device. Plus, how often do you expect to perform this operation? Don't you think you are trying to optimize something that does not have any real performavnce impact here? -- Dmitry