CompuServe Thread

#Rumor Update

6 messages in this thread
#151719From: Thomas HarmonJun 9, 1994 7:39 PM
running simulations on xylinks(or that matter any VLSI) for a decent check will take months or more! (I am a chip weenie, what can I say?, it pays the bills!) Up front specs for a chip has Gotta be done! I have learned the hard way! What we try to do is look at the software algorithm that the do in software, make it as effecient as we can think of, and then make the hardware to do as much to off load the processor as possible.. Thats what we try to do.. Now if a vlsi already exists and the designer left the company and didn't document it is very painful..(the foundary doesn't want to make 1.2u chips any more, could you translate it to fred's at .6u and by the way we would like to add this, delete that and change the other, and yes could you have it done by next friday?) -Tom-
#151725From: Dale LarsonJun 9, 1994 8:04 PM
Sure, there was a new spec up to a point. The problem with the details is how to know what is important — with thousands (tens of thousands) of programs out there depending on all kinds of weirdness, how do you decide what belongs in the spec and what doesn't short of saying "a reasonable sample of this software must run correctly?" At some point, you still have to take the new design and the old design and put them side by side and make sure they work alike. That was the part I was involved in. My job at that time was in testing, not design. BTW, I meant emulation, not simulation — the Xylinks ran in real time. If we found a bug, the chip weanie went an compiled a new netlist or something and downloaded it into the Xylinks angain and we tested again. Sure, it sometimes took a day (lots of other work on those Suns)… — Dale L. Larson, Intangible Assets Manufacturing
#152258From: Thomas HarmonJun 13, 1994 5:50 PM
So you grabbed lots of of the shelf programs and used them as the test bench stimulus on the xylinks.. Ya I can believe weirdness in programming from all those programs.. But what Should have happened was that the original designer of the chip should have spec'ed out what he was building before he built it and that should be the 'master' for the chip.. (not your new chip, the old chip!) Sounds like what you did is a decent method for duping an existing chip.. interesting to hear whats going on, thanks for the info… -Tom-
#152281From: Dale LarsonJun 13, 1994 7:57 PM
You got it (though I also wrote test programs in which I tried to characterize some of the fuzzier and more obscure points). I know what should have happened. Of course, I think the 8520 is based on another CIA which wasn't fully speced and that they are both late 70s. — Dale L. Larson, Intangible Assets Manufacturing
#152511From: Eulogio (DJ) GarciaJun 15, 1994 6:23 PM
There's probably a fair chance the 8520 is a derivative of the 6520 that MOS Technology put out with the 6500 series mp which in itself was a derivative of the 6820 from Motorola … DJ
#152753From: Black Belt SystemsJun 17, 1994 1:09 AM
Actually, I think it's a derivative of the 6522 – there's _very_ little difference between them. –Ben …via AutoPilot