Date: Sun, 5 Oct 1997 05:52:50 -0500 (EST) From: "John S. Dyson" <toor@dyson.iquest.net> To: joerg_wunsch@uriah.heep.sax.de Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: New SCSI Framework Patches Available Message-ID: <199710051052.FAA00587@dyson.iquest.net> In-Reply-To: <19971005115559.AS58851@uriah.heep.sax.de> from J Wunsch at "Oct 5, 97 11:55:59 am"
next in thread | previous in thread | raw e-mail | index | archive | help
J Wunsch said: > > I would prefer a transition from b_blkno to b_offset, thereby > eliminating all the ``shift right here, shift back left there'' mess > completely. The drawback is that b_offset needs to be 64 bits, so > naturally this will lead to more bloat and slower execution on 32-bit > CPUs. The second problem is that b_blkno is being abused in quite a > number of places (see b_lblkno and b_pblkno), making it more work to > do the transition. Of course, the latter can be thought of as an > advantage, in that it requires some code cleanup. ;-) > > The advantage (besides of achieving the goal to support arbitrary > block sizes) is that the code will be cleaner. The 64-bit problem > will no longer be a problem once running on a 64-bit CPU. > SCSI isn't probably the best place to discuss the eventual transition from b_blkno to b_offset, but I don't think that it will cause us very much inefficiency to move to the offset representation. My opinion (guess) is due to all of the conversions that we are already doing being costly. The cost is compounded by the necessity for us to do alot of the calculations in 64bits anyway. If we think that this should be a 3.0 thing, let's make the decision ASAP, and make sure that DG, BDE, and PHK get involved in the discussions. -- John dyson@freebsd.org jdyson@nc.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199710051052.FAA00587>