Path: utzoo!attcan!uunet!husc6!bloom-beacon!tut.cis.ohio-state.edu!mailrus!ames!pasteur!ucbvax!decwrl!nsc!grenley From: grenley@nsc.nsc.com (George Grenley) Newsgroups: comp.sys.nsc.32k Subject: Re: commentary on "532 Manifesto" Keywords: 32532, PD board Message-ID: <5083@nsc.nsc.com> Date: 23 May 88 23:03:32 GMT References: <17145@gatech.edu> <355@myab.UUCP> Reply-To: grenley@nsc.UUCP (George Grenley) Organization: National Semiconductor, Sunnyvale Lines: 65 This is an interesting discoussion, but it is missing a couple of pertinent facts... Regarding the byte swapping issue, NSC's VME532 (and 332) boards "swap" the bytes, i.e., we put byte 0 where a 68xxx would expect to find it. This way test files transfer from 68xxx Unix (or whatever) machines to 32000 machines with no hassle. Obviously you want source level compatibility. The binaries come out "wrong", but we couldn't run 68000 binaries any way, so no loss there.... In article <355@myab.UUCP> lars@myab.UUCP (Lars Pensj|) writes: >In article <17145@gatech.edu> ken@gatech.edu (Ken Seefried iii) writes: >> >> Daryl McDaniel mentioned the fact that he has been designing '532 >>systems with the AT&T 32104 DMA chip. The more i look at the specs for the >>32104, the better this sounds. I vote for DMA. I really think for a chip >>with the speed of the '532, we need it. >We have considered designing a 532-system with 32104 DMA. However, there are >problems with this chip: Heurikon is about to release their 52532 based VME board, which uses the AT&T device. I don't know if that's where Daryl McDaniel works or not, but if it isn't perhaps some Heurikon engineer on the net would care to comment... >1. It has another byte order than 532. That means that you have to swap > the bytes in the system bus interface (nobody want tapes with swapped bytes). See above discussion re byte swap, go get pencil and paper, and see for your self. I can't expalin it in text, you need to look at a diagram. Also, consider that the DMA is only moving the data, not trying to interpret it, so it doesn't care about byte order. > Now the software have to swap bytes in addresses. Ugly (and some overhead) ! We haven't found the overhead to be any problem. If ou put DMA on the same "side" of the byte swap as the CPU, it will see the pointers the same. >Now I wonder: are there really no DMA chip which do not have these problems ? >-- > Lars Pensj| > {decvax,philabs}!mcvax!enea!chalmers!myab!lars There is one - "SW DMA". Seriously, the advantage of a DMA exists ONLY in systems where the DMA and the CPU can operate simultaneously. If you have to stop the CPU while DMA runs, you have gained nothing - remember, the 532 can move 4 bytes in 132 nsec at 30 mhz (R+W) - faster than DRAM anyway. So if you are building a low cost system, you don't need a DMA. Heurikon's board is aimed a performance conscious users, and since VME is multi master, and their on-board DRAM is dual port, it makes sense to have DMA - the 532 can still run, with some waitstates, maybe, but still running. A good choice for Unix and specialized real time applications. So why doesn't NSC have a DMA on our VME board? Easy - we couldn't fit it on. We have two boards FULL of chips already..... 'nuff George Grenley NSC (408) 721-5513... call me and buy a board! disclaimer: I hate IBM mainframes, especially PROFS.