Different Implementations of VLIW .
|
Prev: The Mark Thorson FAQ: Read all about this LIAR, BULLY, THUG, CYBERSTALKER and HARASSER
Next: The Mark Robert Thorson FAQ: Read all about this LIAR, BULLY, THUG, CYBERSTALKER and HARASSER
From: Quadibloc on 10 Apr 2008 12:34 On Apr 9, 3:17 pm, n...(a)cus.cam.ac.uk (Nick Maclaren) wrote: > In article <90aLj.47758$kN5.39...(a)newsfe1-gui.ntli.net>,"Wilco Dijkstra" <Wilco_dot_Dijks...(a)ntlworld.com> writes: > > |> "Ripunjay Tripathi" <ripunjay.tripa...(a)gmail.com> wrote in message > |>news:810bb078-2719-48ae-b47a-6510fa72a9bb(a)u12g2000prd.googlegroups.com.... > |> On Apr 9, 12:21 am, Terje Mathisen <terje.mathi...(a)hda.hydro.com>|> wrote: > > |> > Ripunjay Tripathi wrote: > |> > > Different Implementations of the same VLIW architecture may not be > |> > > binary-compatible with each other. > |> > > |> > > I am looking for explaination on the above line. > |> > |> Read thishttp://en.wikipedia.org/wiki/VLIWfor a basic explanation. > > Of which the first sentence is so misleading as to be effectively > false. I can't be bothered to read further. VLIW does take advantage of instruction-level parallelism, just not between different instructions the way decoupled microarchitecture does; it forces you to code it by hand. Or maybe you mean something else. John Savard
From: Quadibloc on 10 Apr 2008 12:39 On Apr 10, 4:26 am, n...(a)cus.cam.ac.uk (Nick Maclaren) wrote: > The point is that VLIW originated as a way of making direct use of > separate hardware units, and was quite commonly used for microcode > (e.g. on the IBM System/370 range). It was 'rediscovered' in the > 1990s as a way of making use of the 'available' ILP in programs > written in 3GL languages (like Fortran or C). > > Wikipedia has got that arsy-versy. The term VLIW wasn't coined to name horizontal microcode, it was coined to name things like the Control Data Cyberplus. The two were still very similar in effect, though, because neither is produced on the fly, so neither can take advantage of ILP between threads the way decoupled microarchitecture can. John Savard
From: Nick Maclaren on 10 Apr 2008 12:56 In article <1192aadf-4742-4c5c-af6a-4e4767e0052d(a)k10g2000prm.googlegroups.com>, Quadibloc <jsavard(a)ecn.ab.ca> writes: |> |> > The point is that VLIW originated as a way of making direct use of |> > separate hardware units, and was quite commonly used for microcode |> > (e.g. on the IBM System/370 range). It was 'rediscovered' in the |> > 1990s as a way of making use of the 'available' ILP in programs |> > written in 3GL languages (like Fortran or C). |> > |> > Wikipedia has got that arsy-versy. |> |> The term VLIW wasn't coined to name horizontal microcode, it was |> coined to name things like the Control Data Cyberplus. The two were |> still very similar in effect, though, because neither is produced on |> the fly, so neither can take advantage of ILP between threads the way |> decoupled microarchitecture can. Yes. But to say that horizontal microcode and the control processors that used what-was-later-to-be-called-VLIW instruction sets were not VLIW because the term hadn't been coined by then is the the sort of terminological politics with which I will have no truck. If the term VLIW means anything useful in computer architecture, those were VLIW. Wikipedia starts "... VLIW refers to a CPU architecture designed to take advantage of instruction level parallelism (ILP)." As I said earlier, that puts the cart before the horse and rewrites history. Regards, Nick Maclaren.
From: Joe Pfeiffer on 10 Apr 2008 13:59 nmm1(a)cus.cam.ac.uk (Nick Maclaren) writes: > In article <1b8wzmiqb2.fsf(a)babs.wb.pfeifferfamily.net>, > Joe Pfeiffer <pfeiffer(a)cs.nmsu.edu> writes: > |> nmm1(a)cus.cam.ac.uk (Nick Maclaren) writes: > |> > In article <90aLj.47758$kN5.39126(a)newsfe1-gui.ntli.net>, > |> > "Wilco Dijkstra" <Wilco_dot_Dijkstra(a)ntlworld.com> writes: > |> > > |> > |> Read this http://en.wikipedia.org/wiki/VLIW for a basic explanation. > |> > > |> > Of which the first sentence is so misleading as to be effectively > |> > false. I can't be bothered to read further. > |> > |> Only when taken completely out of context. > > Not at all, but you probably haven't been in this game long enough to > realise why I said that. To start an article by putting the cart > before the horse and rewriting history isn't good, even in Wikipedia, > though it is regrettably common. I certainly remember horizontal microcode. You might be interested in re-reading Fisher's VLIW paper, which the wikipedia article you find so misleading contains a link to, and which does a very good job explaining the similarities and differences between VLIW and horizontal microcode, and how his goal was specifically to improve ILP.
From: Nick Maclaren on 10 Apr 2008 15:12
In article <1b7if5vbfc.fsf(a)babs.wb.pfeifferfamily.net>, Joe Pfeiffer <pfeiffer(a)cs.nmsu.edu> writes: |> |> > Not at all, but you probably haven't been in this game long enough to |> > realise why I said that. To start an article by putting the cart |> > before the horse and rewriting history isn't good, even in Wikipedia, |> > though it is regrettably common. |> |> I certainly remember horizontal microcode. You might be interested in |> re-reading Fisher's VLIW paper, which the wikipedia article you find |> so misleading contains a link to, and which does a very good job |> explaining the similarities and differences between VLIW and |> horizontal microcode, and how his goal was specifically to improve ILP. I wasn't referring just to horizontal microcode, but to the control systems that used VLIW instruction sets. I haven't read Fisher's paper, mainly because the abstract set off my, er, aromatic hyperbole alarms so loudly. That was because I remembered the research into VLIW as a target in the 1960s or 1970s, and why those people decided it was a non-starter. In particular, his 'expectation' of a 10- to 30-fold improvement by the use of what sounded like a variation of an old technique would imply that you could use the same methodology to get a 3- to 10-fold improvement on more conventional hardware. I may be maligning him, but EPIC assuredly hasn't delivered on what his abstract said he expected, by a factor of 10-30, so I suspect my gut feeling was not too far off. Regards, Nick Maclaren. |