Path: csiph.com!usenet.pasdenom.info!weretis.net!feeder4.news.weretis.net!news.musoftware.de!wum.musoftware.de!fu-berlin.de!uni-berlin.de!individual.net!not-for-mail From: "David Holden" Newsgroups: comp.sys.acorn.misc Subject: Re: Unipod query Date: Thu, 26 Jul 2012 08:34:37 GMT Organization: APDL Lines: 32 Message-ID: References: <52b09af13ccvjazz@waitrose.com> Reply-To: SpamBin@apdl.co.uk X-Trace: individual.net J6cbZX2b3EngYLheAFkQRgkdfXXyD/P3s5vqOJswQ1FrVuj8+68UKcSe1r4AMPjKje Cancel-Lock: sha1:zZGoRN0Sd0dkYPhhz2+55PJW1dQ= X-Newsreader: News Rover 12.3.0 (http://www.NewsRover.com) Xref: csiph.com comp.sys.acorn.misc:5937 On 25-Jul-2012, Tony Moore wrote: [results not reproduced for brevity] > The first group of tests show that, for byte access, IDEFS is, on > average, more than 60% _slower_ than ADFS. > > The second group of tests show that, for load/save operations, IDEFS is, > on average, about 50% faster than ADFS. The byte access speed increase is well known and is because ADFS has buffered byte read/write. With our IDE interfaces we did address this with a version which had buffering and this was a bit faster than ADFS. However this caused problems with high speed access, and proved incompatible with DMA, so we reverted to unbuffered in/out. This gives maximum data transfer for larger files, i.e. 95% of stuff, and if someone is using a program which does a lot of byte access and which they use frequently then if maximum speed is important just put the data files for that program on an ADFS disc. We did make a special buffered version with no DMA available at one time for people for whom this was the primary requirement. I presume similar problems were encountered by Simtec with the Unipod. BTW the results I get for larger file transfer on a Unipod are quite a bit faster than the ones you show. -- David Holden - APDL -