From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: [PATCH v2 1/3] scsi_cmnd: Introduce scsi_transfer_length helper Date: Thu, 26 Jun 2014 12:38:42 -0400 Message-ID: References: <1402477799-24610-1-git-send-email-sagig@mellanox.com> <1402477799-24610-2-git-send-email-sagig@mellanox.com> <53AC3402.2080302@acm.org> <1403794551.3572.6.camel@dabdike> <86EE1A5C-8943-4C87-9A79-899DE85971CA@ornl.gov> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Return-path: In-Reply-To: <86EE1A5C-8943-4C87-9A79-899DE85971CA@ornl.gov> Sender: linux-scsi-owner@vger.kernel.org To: "Atchley, Scott" Cc: Bart Van Assche , Sagi Grimberg , "" , "" , "" , "" , "" , "" , "" List-Id: linux-rdma@vger.kernel.org On June 26, 2014 11:41:48 AM EDT, "Atchley, Scott" wrote: >On Jun 26, 2014, at 10:55 AM, James Bottomley > wrote: > >> On Thu, 2014-06-26 at 16:53 +0200, Bart Van Assche wrote: >>> On 06/11/14 11:09, Sagi Grimberg wrote: >>>> + return xfer_len + (xfer_len >> ilog2(sector_size)) * 8; >>> >>> Sorry that I just noticed this now, but why is a shift-right and >ilog2() >>> used in the above expression instead of just dividing the transfer >>> length by the sector size ? >> >> It's a performance thing. Division is really slow on most CPUs. >> However, we know the divisor is a power of two so we re-express the >> division as a shift, which the processor can do really fast. >> >> James > >I have done this in the past as well, but have you benchmarked it? >Compilers typically do the right thing in this case (i.e replace >division with shift). The compiler can only do that for values which are reducible to constants at compile time. This is a runtime value, the compiler has no way of deducing that it will be a power of 2 James -- Sent from my Android device with K-9 Mail. Please excuse my brevity.