as@aipna.ed.ac.uk (01/08/91)
> Is this true? I'm not sure how you define Interrupt Latency > ... the literature ... tells me that... > Fast IRQ takes 12 cycles and RTI from it takes 6 cycles. > NMI or IRQ takes 21 cycles and RTI from it takes 15 cycles. > SWI takes 19 cycles and RTI from it takes 15 cycles. > SWI2 and SWI3 take 20 cycles and RTI from them takes 15 cycles. Yep, I think these are the figures, I made a similar point to the Acorn eng., but he seemed pretty adamant about it. Ditto other Acorn types I came across and winged at at the time. I personally think the reasons for going with the 6502 were more socio-political than well-reasoned out. I.e. 6502's were cheap, and Acorn's programmers were mainly 6502 assembler junkies. Once they'd made their minds up I'm sure all kinds of good reasons could be found why the 6809 couldn't hack it. Their other rather questionable decision taste-wise wise was that weird infatuation with BCPL... (Built in Cambridge Programming Language as the wags would have it). > (GWP - Gods Wonderful Processor) Wasn't it just - think how nice stuff like paged ROMs etc could have worked with a changeable base page and 16 bit indexed modes. I daren't think of a system with OS/9 and the 6809 MMU. I guess after they spent all the board real estate on speech processors, teletext support, silly tube connector etc, there wasn't much budget left for the CPU. -- Andrew Stevens, JANET: as@uk.ac.ed.aipna Dept. of Artificial Intelligence, ARPA: as@aipna.ed.ac.uk 80 South Bridge, UUCP: ...!mcvax!ukc!aipna!as Edinburgh University, EDINBURGH
gtoal@tharr.UUCP (Graham Toal) (01/09/91)
>> (GWP - Gods Wonderful Processor) > >Wasn't it just - think how nice stuff like paged ROMs etc could have worked >with a changeable base page and 16 bit indexed modes. I daren't think of >a system with OS/9 and the 6809 MMU. I guess after they spent all >the board real estate on speech processors, teletext support, >silly tube connector etc, there wasn't much budget left for the CPU. Actually I built a 6809 2nd processor while I was working at acorn -- rather neat though i say so myself -- it was a 4 package tiny board -- custom ram address decoder/refresher, rom, 64Kram sil & cpu. I was hacking VLSI at the time and it was for a test chip (it worked), but I never got as far as putting any real software on it. It would have gone fast enough for tube operation using the FIQ and NMI intrrupts slightly differently from the obvious way Acorn considered when they looked at it and said it was too slow :-) Actually I was a bit depressed by it, because to prototype the software I wrote a 6809 emulator for the 32016 Panos machine... and it ran faster than the real chip :-( Motorola never put the hardware engineers onto speeding it up beyond 2Mhz unlike the Z80 or 6502 rivals. Probably because it had been designed by software people in the first place :) I still have it by the way, if any real fanatic is serious about getting it running. You'd have to know the tube code inside out and know a lot of electronics to do a bit of necessary circuit redesign. Graham -- (* Posted from tharr.uucp - Public Access Unix - +44 (234) 261804 *)