richard@gryphon.COM (Richard Sexton) (04/29/89)
This is a flame. It has technical content relevent to comp.sys.amiga.tech but it's a rehash. If you understood how this stuff worked the first time, skip the rest of this article. If you don't like flames, skip this article. Proceeding any further removes your right to coplain about this being a flame. you all know how I hate to flame, but JESUS, STEVE, did you read anything I wrote ? In article <33756@kilowatt.uucp> raz@sun.UUCP (Steve -Raz- Berry) writes: >In article <15147@gryphon.COM> richard@gryphon.COM (Richard Sexton) writes: >)In article <33739@kilowatt.uucp> raz@sun.UUCP (Steve -Raz- Berry) writes: > >))I think we are a little confused here, There is no way that the Weitek >))can possibly emulate a 68881. Not without a major hardware kluge. I think >))that richard means that the Weitek chip SET (IU and FPU) together function >))simalirly to the 680X0 and coprocessor series. > >)No, Richard meant it was a co-processor, like the 68881/2. Thats what >)the ads implied. > >Your kidding. I don't claim to know all, but I thought I woulda >heard that. Oh well. > >)So I called Weitek. > >)Buggers. > >)The stupid thing is memory mapped. You write the operands into >)memory addresses, then give it an operation, then poll it for >)completion. Some co-processor. > >Not a true co-processor if you ask me, ok so it fits the description... >but it's a kluge. No Steve. A co-processor would execute math instructions out of the 680x0 instruction stream. This thing acts more like a peripheral chip in that you have to write operands and what operation to perform to the chip; poll for completion, then read the result. But I said that. >)Some performence data: > >)Linpack - single precision 8 MFLOPS >) double 6 MFLOPS >)Whenstone Single precision 2.0 MWhets (is that what he said ? Mwhets ?) >) Double 1.2 > >Did it say how this was set up? What I mean is, is it a 680x0 peripheral >or is it running it's own code memory (probably). If it needs it's own >special memory, like cache, then you might as well build a seperate board >and do it up as black box math server. If you are going to go to that >much trouble, might as well go with the '860 and get REAL speed. >I guess I define co-processor as a transparent hardware accelerator. >If you can hook it up to the main processor, with minimum amount of Thats ``Hassle''. I hate seeing good hippie words misspelt. The rest I don't care about. >hassel, and have it run out of processor memory space, then it's a >co-processor. My definition. Weitek would probably disagree. Steve, get a grip. The 6888x is a true co-processor. Of course your definition, while, uhh, *unique* is not the one generally held by the rest of the world. This Weitek thing is a peripheral math processor, which thay are *thinking* about making into a single chip if enough people call them and say it's a great idea, we'll take 100,000. But it'll still be another peripheral chipm not something that executes 680x0 floating point instructions. But I said that. >)How do these figures compare with '881/2 and 860 ? > >I have no clue. I looked in the 881 manual and couldn't find the performance >figures. Thats nice Steve, but I don't believe you. I'm not going to believe it until EVERYBODY in this group posts that they dont know. Did I mention that I don't know the spin on a fermion ? >)Some interesting stuff about this part - it doesnt exist. It seems >)Weitek has a nice math chip, the 3116 or something like that. They >)sell a daughtorboard that has some glue logic and this chip >)for use with the 386. What they have for the 68000 is a board >)that uses the 3116 and some glue. > >Is it compatable with the 881 instruction set? If not, why bother? You'ld >have to write your own compiler for it. YECHH. Unless it does sin(x)/cos(y) >in two cycles... No Steve, for the third time, it doeesnt execute 680x0 floating point instructions. It's a fucking peripheral chip. You don't have to write your own compiler for it, you talk to it through memory mapped addresses. Many of us have found it possible to do this without re-writing a compiler. And no it doesnt do sin(x)/cos(y) in two cycles. It doesn even do a divide in two cycles. I asked if this thing was fast enough to not need polling for completion, like the old 4Mhz Z-80/9511 setups. He said: ``Well, some instructions take longer and you need to poll for completion''. ``Oh, lke which ones'' I asked. ``Like divide'' he said. Oh. >)Every time I pumped him for information, he pumped me >)wanting to know how many i needed, so thay could make their >)projections and deicide to make this chip or not. > >Should have asked for a sample ;-) Of what ? Their board ? Dont be a dork. It isn't a chip yet. But I said that. >)Hell, I told 'em 10,000. Couldnt hurt. Weitek is in Sunnyvale. > >Gee, I can pick one up on my way to work! > >I don't like Weitek. When I was at Raster Tech. I talked to a few of the >engineers that had worked on previous projects involving their 64 bit >FPU/IU chip set. I heard nothing but horror stories about hardware bugs >and instructions that didn't work as advertized. Abort and stall are good >examples. Oh, well thats it then. Steve Berry doesnt like Weitek. They may as well close up shop right now. I'm calling my stockbroker right now to tell him. So use the instructins that do work, dipshit. The Weitek part on a (IBM) PC gets about 3X the equiv 'x87 performence. >Steve -Raz- Berry Disclaimer: I didn't know nutin! Oh, that explains it. And theres two t's in nuttin'. Arrrg. Why do I bother? -- "My latest 'problem', btw, is that I'm working out an opportunity to get laid with some girl over the net." - Ted Kaldis richard@gryphon.COM decwrl!gryphon!richard gryphon!richard@elroy.jpl.NASA.GOV
raz@kilowatt.uucp (Raz- Berry) (05/01/89)
In article <15248@gryphon.COM> richard@gryphon.COM (Richard Sexton) writes:
)you all know how I hate to flame, but JESUS, STEVE, did you read anything
)I wrote ?
Sure, it was a little confused, but I read it.
)))The stupid thing is memory mapped. You write the operands into
)))memory addresses, then give it an operation, then poll it for
)))completion. Some co-processor.
))Not a true co-processor if you ask me, ok so it fits the description...
))but it's a kluge.
)No Steve. A co-processor would execute math instructions out of the
)680x0 instruction stream. This thing acts more like a peripheral chip
)in that you have to write operands and what operation to perform
)to the chip; poll for completion, then read the result. But I said that.
Essentially a co-processor just means 'in conjunction' with. The COPPER
in the Amiga has to be fed instructions just like a peripheral chip.
You build a Copper list in memory and then tell the copper sequencer
(by slamming it's code address into a dedicated register) where to start
executing it's next instruction stream.
You can do the same kind of thing with the Weitek part, if it's smart enough.
But by your definition, the Amiga Copper is misnamed. Gee, I guess that means
Commodore will have to rewrite all their hardware manuals.
But basicly I was agreeing with you, why get nasty?
)))Some performence data:
))
)))Linpack - single precision 8 MFLOPS
))) double 6 MFLOPS
)))Whenstone Single precision 2.0 MWhets (is that what he said ? Mwhets ?)
))) Double 1.2
))
))Did it say how this was set up? What I mean is, is it a 680x0 peripheral
))or is it running it's own code memory (probably). If it needs it's own
))special memory, like cache, then you might as well build a seperate board
))and do it up as black box math server. If you are going to go to that
))much trouble, might as well go with the '860 and get REAL speed.
))I guess I define co-processor as a transparent hardware accelerator.
))If you can hook it up to the main processor, with minimum amount of
)
)Thats ``Hassle''. I hate seeing good hippie words misspelt. The rest
)I don't care about.
Well, I guess I should just curl up and die eh? I feel like I am tress-
passing in some tigers lair, and if I don't agree with what the idiot
tiger thinks, or I say things a little differently... tiger bait.
I was looking for information as to how they got their performance #'s.
Good answer Rich.
))hassel, and have it run out of processor memory space, then it's a
))co-processor. My definition. Weitek would probably disagree.
)Steve, get a grip. The 6888x is a true co-processor. Of course
No shit. I was talking about your fabled Weitek peice of crap.
I *know* it is not a "true" co-processor. I *know* it is a polled device.
My point was if you can't just plug'n play with the 68881 without
significant speed advantages, then it is not worth the silicon it's
cut from.
)your definition, while, uhh, *unique* is not the one generally
)held by the rest of the world. This Weitek thing is a peripheral
Ahh, the world speaks.
)math processor, which thay are *thinking* about making into a single
)chip if enough people call them and say it's a great idea,
)we'll take 100,000. But it'll still be another peripheral chipm
)not something that executes 680x0 floating point instructions.
)But I said that.
No, you didn't.
)))How do these figures compare with '881/2 and 860 ?
))I have no clue. I looked in the 881 manual and couldn't find the performance
))figures.
)Thats nice Steve, but I don't believe you. I'm not going to believe
)it until EVERYBODY in this group posts that they dont know.
Gee, I'm hurt. Try looking in an applications note dummy, I can't do everything
for you.
)))Some interesting stuff about this part - it doesnt exist. It seems
)))Weitek has a nice math chip, the 3116 or something like that. They
)))sell a daughtorboard that has some glue logic and this chip
)))for use with the 386. What they have for the 68000 is a board
)))that uses the 3116 and some glue.
If all they have is a board, then who's machine does it plug into?
Not an Amiga I bet. Or is it a daughter board like the 386 counterpart?
Try to be a little clearer richard, not everyone is an ESPer like you.
))Is it compatable with the 881 instruction set? If not, why bother? You'ld
))have to write your own compiler for it. YECHH. Unless it does sin(x)/cos(y)
))in two cycles...
)No Steve, for the third time, it doeesnt execute 680x0 floating
)point instructions. It's a f***ing peripheral chip. You don't have
)to write your own compiler for it, you talk to it through memory
)mapped addresses. Many of us have found it possible to do this
)without re-writing a compiler. And no it doesnt do sin(x)/cos(y)
)in two cycles. It doesn even do a divide in two cycles. I asked
If you don't have compiler support, this severly limits the number of
people who would be interested in this kluge. Except you Richard, I'm
sure you will be perfectly happy doing POKE valx, 2104948.
Wait, does it being a peripheral chip imply instruction set incompatability?
Interesting design concept. Just because you memory map something
doesn't mean it can't execute instructions compatible with the native
coprocessor. It doesn't mean you can't microprogram it like the TI
graphics chip, although I admit for a math chip this would be unusual.
)if this thing was fast enough to not need polling for completion,
)like the old 4Mhz Z-80/9511 setups. He said: ``Well, some instructions
)take longer and you need to poll for completion''. ``Oh, lke which
)ones'' I asked. ``Like divide'' he said. Oh.
You mean you didn't call him dork or dipshit?
)))Every time I pumped him for information, he pumped me
)))wanting to know how many i needed, so thay could make their
)))projections and deicide to make this chip or not.
))Should have asked for a sample ;-)
)Of what ? Their board ? Dont be a dork. It isn't a chip yet. But
)I said that.
What's the matter Richard? Ever hear of engineering samples? If they
are talking about the part, then they *should* have samples to give to
the designers to use to build proof of concept boards. If they don't
have anything to give you, then they are not for real. Besides they would
have at least sent you a 3116 with an ap note on how to interface it to
your favorite processor.
))I don't like Weitek.
)Oh, well thats it then. Steve Berry doesnt like Weitek. They may as
)well close up shop right now. I'm calling my stockbroker right
)now to tell him.
)So use the instructins that do work, dipshit. The Weitek part
)on a (IBM) PC gets about 3X the equiv 'x87 performence.
Christ.
On an IBM PC that may be just peachy. When you are building a board that
needs absolute speed, runs it's own code concurrent with the host processor
to do dedicated Goureau (sp??) shading of 200K 3D polygons/sec, the last thing
you need to deal with is vendor bugs. Especially when they know about them
in advance and conviently forget to inform you.
)Oh, that explains it. And theres two t's in nuttin'.
Clever. Which stooge were you?
)richard@gryphon.COM decwrl!gryphon!richard gryphon!richard@elroy.jpl.NASA.GOV
This discussion has gone beyond anything remotely Amiga specific. I suggest
that if you (yes you Richard) are interested in continuing the thread that
we move to another news group, or e-mail. Based on the tone of this flame,
it's hard for me to believe that anything of interest is forthcoming.
--
Steve -Raz- Berry Disclaimer: I didn't do nutin!
UUCP: sun!kilowatt!raz ARPA: raz%kilowatt.EBay@sun.com
"Fate, it protects little children, old women, and ships named Enterprize"richard@gryphon.COM (Richard Sexton) (05/03/89)
I'm on a wierd diet.
Pickled onions and pineapples would do that to anybody.
--
``The way to heaven is through weasel lore!'' - Ted Kaldis
richard@gryphon.COM decwrl!gryphon!richard gryphon!richard@elroy.jpl.NASA.GOV