VLIW pre-history
|
From: Eric Smith on 17 May 2007 22:26 Quadibloc <jsavard(a)ecn.ab.ca> writes: > There are HARVEST manuals on Al Kossow's site now. I haven't been able > to make enough sense out of them to have a definite opinion, but it is > at least possible for this machine to be similar to a dataflow > architecture. The Harvest coprocessor was sort of equivalent to wiring plugboards on unit-record equipment, but with a little more sophisticated capability than such equipment. It basically processed streams of bytes, doing table lookups and the like. It's VLIW-like in the sense that the instruction has lots of fields for controlling different aspects of the operation of the processor (not quite like independent functional units, though). However, it has no program counter, and executes a single instruction (set up by the main processor), which may take arbitrarily long to complete since it processes an arbitrarily long stream of data.
From: Quadibloc on 18 May 2007 01:37 Eric Smith wrote: > It's VLIW-like in the sense that the instruction has lots of fields > for controlling different aspects of the operation of the processor > (not quite like independent functional units, though). However, > it has no program counter, and executes a single instruction (set up > by the main processor), which may take arbitrarily long to complete > since it processes an arbitrarily long stream of data. .. Whereas the real classic VLIW machines used the fact that they *did* have a program counter to use a short loop of instructions to, in a sense, simulate a computer with a larger number of functional units. I'll agree that this isn't true VLIW, just "vaguely VLIW-ish", as you've put it. But with a dataflow architecture, the program counter is a means and not an end, so the relationship may be closer than it seems. John Savard
From: Alan Charlesworth on 18 May 2007 09:08 > I'm curious if Alan Charlesworth (who posted in this thread > as well) had any contact with the i860 designers wrt to the > FP unit? Or did the designers borrow the ideas independently? > Does Les Kohn lurk here, or Sai Wai Fu? I've always wanted > to know the design details-- the IEEE Spectrum article in 1989 > talks more about staffing, as opposed to the technical details. > Nope. I stayed at FPS until it went under and its assets were bought by Cray in 1992. Those assets later turned into the Starfire E10K, which were sold to Sun when SGI bought Cray in 1996 -- but that is a different story. As for the i860, FPS used in a matrix coprocessor board, which was an add-in to the FPS-164, a 64-bit follow-on to the AP-120B.
From: Eugene Miya on 18 May 2007 13:06 In article <1179452578.350788.201940(a)h2g2000hsg.googlegroups.com>, Quadibloc <jsavard(a)ecn.ab.ca> wrote: >> On what basis do you think it might be? > >There are HARVEST manuals on Al Kossow's site now. I haven't been able >to make enough sense out of them to have a definite opinion, but it is >at least possible for this machine to be similar to a dataflow >architecture. I didn't see Al in his cube on Wednesday (I did specifically look over there, but his chair was empty). I've no time to read ancient Harvest manuals. I do know something about dataflow architectures. Static or dynamic DF? Some of Jack's ideas are antithetical to VLIW and Arvind might take some issue as well. --
From: Eugene Miya on 18 May 2007 13:12
In article <qhfy5urhwb.fsf(a)ruckus.brouhaha.com>, Eric Smith <eric(a)brouhaha.com> wrote: >eugene(a)cse.ucsc.edu (Eugene Miya) writes: >> My position is that I don't regard of these pre-Trace machines as VLIW. >> But go after John. Actually N. who programmed on the Harvest, I saw >> yesterday. > >The Harvest system (IBM 7950) is a superset of the Stretch system >(IBM 7030). Yea yea, that I know. >The Stretch processor (IBM 7101 Central Processing Unit) was not >at all like a VLIW. Yea yea, that I know. >The Harvest coprocessor (IBM 7951 Processing Unit) was very unusual >and perhaps vaguely VLIWish. On what basis? It is true that the first Trace I saw was at the Harvest's owner. I have a little idea of what they do on the basis of at least 3-4 of the research architectures they did on the basis on Harvest. It's also true they have machines I know nothing about, one series being built in the very city in which I live, others near where I grew up. Using the Trace as a standard, how is the 7951 uniquely like the Trace? I can draw comparisons to the CM, the Maspar and the MPP. That's easy. -- |