aek@Apple.COM (Al Kossow) (06/19/88)
In the bad old days when I did hardware, sometimes it's a real pain to make registers readable. Take the case of a counter-chain made up of lots of 4bit up-down counters. None of the counter outputs were bussed together, so it was necessary to have lots of tri-state buffers or n to 1 multiplexers going back to the data bus which would only used to read the output values of the counters. Sometimes when you're running tight on pcb (or chip) real-estate you have to weigh how important it is to have lots of area tied up in space-hogging buses just for reading a register back. -- Al Kossow @ Apple Computer, Inc., Cupertino, CA UUCP: {mtxinu,sun,nsc,voder}!apple!aek ARPA: aek@apple.apple.com
crowl@cs.rochester.edu (Lawrence Crowl) (06/20/88)
In article <12441@apple.Apple.COM> aek@Apple.COM (Al Kossow) writes: >Sometimes when you're running tight on pcb (or chip) real-estate you have to >weigh how important it is to have lots of area tied up in space-hogging buses >just for reading a register back. This is fine so long as you evaluate the SYSTEM costs and benefits (including reliability). These include software development and maintenance costs as well as hardware costs. The reason software costs are so often neglected in this analysis is that it is difficult to estimate the impact of hardware features on software costs. However, the $100,000/yr cost for a software engineer will pay for an awful lot of hardware in anything but consumer or commodity products. -- Lawrence Crowl 716-275-9499 University of Rochester crowl@cs.rochester.edu Computer Science Department ...!{allegra,decvax,rutgers}!rochester!crowl Rochester, New York, 14627
nather@ut-sally.UUCP (Ed Nather) (06/20/88)
In article <12441@apple.Apple.COM>, aek@Apple.COM (Al Kossow) writes: > > Sometimes when you're running tight on pcb (or chip) real-estate > you have to weigh how important it is to have lots of area tied up in > space-hogging buses just for reading a register back. > As a matter of interest, is this also true when the original specification demands that registers be readable, or is it primarily because the rest of the design is already done, and the read process becomes an "add-on" that is costly? I realize this question is application-dependent, but are there some applications where an early awareness of the requirement might lead to a somewhat different architecture that could accomodate read-back in a more economical manner? -- Ed Nather Astronomy Dept, U of Texas @ Austin {allegra,ihnp4}!{noao,ut-sally}!utastro!nather nather@astro.AS.UTEXAS.EDU
baum@Apple.COM (Allen J. Baum) (06/21/88)
[] >In article <12441@apple.Apple.COM>, aek@Apple.COM (Al Kossow) writes: > > Sometimes when you're running tight on pcb (or chip) real-estate > you have to weigh how important it is to have lots of area tied up in > space-hogging buses just for reading a register back. > I've designed hardware both ways. In my "youth" I designed systems that had write only registers. It saved chips. We didn't build very many of the beasts, but they were moderately complex. I had to build them, debug them, and write diagnostics for them. I immediately started believing that the registers should be read/write. Debugging and diagnostics were hell to write. It wouldn't have made that big a difference to our budget; we didn't build very many, after all. But it would have made the project go a much faster, and thats worth a LOT of chips. On the other hand, I've built cards that I just couldn't afford the extra chips to do the readback. It was a fairly simple board (20 chips+RAM), and lots were built. I've never felt bad about the tradeoff- the board was simple enough that there wasn't that much to test anyway, and simple enough that the write only register could be worked around and tested. I've also built a more complex board, with its own microprocessor and self-test diagnostics, and made sure that the microprocessor could read every bit of state and every register in the machine. It was a godsend, and I haven't regretted that decision either. So, sometimes you can get away with it, and sometimes you can't; you just have to know when the right time is. -- {decwrl,hplabs,ihnp4}!nsc!apple!baum (408)973-3385