Why so little parallelism?
|
From: BDH on 2 Nov 2006 18:33 > > > Programs where it is KNOWN to be infeasible > > > > What examples of these do you like? > > 10 print "hello" : goto 10 How about something useful and terminating? > wouldn't the ideal language look a bit like a spreadsheet, with each > cell having formula, whose evaluation is triggered by any value cell it > depends on being written? No need with a decent graph theoretic flow analysis to suss out time dependencies. > p.s. i have not used much apl, but am a registered J user, but to be > honest, forth is my current usage. K is ok but J...well I would say J tends to introduce complicating abstractions with low returns on effort.
From: Andrew Reilly on 2 Nov 2006 18:40 On Thu, 02 Nov 2006 02:42:26 -0800, BDH wrote: >> How would your language be >> different from Verilog or VHDL or Occam? What would be different, this >> time? > > Or are you asking what I think of locks and message passing and so > forth? Well, I'm more concerned with VLIW type techniques. Do VLIW type techniques (plausibly) scale to anything like the degree of parallelism available at the moment with locks and messages (tens of thousands of FUs)? Isn't communication and memory infrastructure still the main issue/concern? Cheers, -- Andrew
From: BDH on 2 Nov 2006 18:48 > >Unfortunately neither you nor mr fuller have come up with superior > >solutions for many of these problems. Domes haven't replaced > >rectilinear frame construction, and alternatives haven't replaced von > >neumann. > > Not only that. Stewart Brand recanted his support for domes in his smart > buildings book. This is not to say that they don't use uses like > radomes. And it was once useful "largest dome in Livermore" to find a > future officemate's house. Boy what a time to leave my RSA data frobb at > home.... No access to my quote database for what Brand said. Eh, I'm not trying to sell domes here.
From: Richard on 2 Nov 2006 18:49 [Please do not mail me a copy of your followup] eugene(a)cse.ucsc.edu (Eugene Miya) spake the secret code <454a4a42$1(a)darkstar> thusly: >eugene(a)cse.ucsc.edu (Eugene Miya) spake the secret code > no secrets ><45491e94$1(a)darkstar> thusly: >>>As a start, you should read Amdahl's (law) paper. It's barely 3 pages >>>long. Amdahl himself has given permission and it's the comp.parallel >>>FAQ panel on the 20th at the very end. It goes out monthly. > >In article <eib7hd$hhk$1(a)news.xmission.com>, Richard <> wrote: >>Also: <http://en.wikipedia.org/wiki/Amdahl's_law> > >Whereas this is not a bad description (I appreciate what Jimmy has done >with wikipedia to its limits), it's not the original work. [...] I tossed that link out there because I tried to find it in the comp.parallel FAQ and while the copy I was viewing was formatted oddly (looks like a bad brute force ASCII -> HTML conversion), I couldn't find Amdahl's paper in there. Got a link to the paper proper? -- "The Direct3D Graphics Pipeline" -- DirectX 9 draft available for download <http://www.xmission.com/~legalize/book/download/index.html>
From: BDH on 2 Nov 2006 18:58
> Well other than an opinion about one language, you haven't refuted > anything about object oriented programming. No matter which way you > try to state it, its a stupid and baseless argument. There is no commonly accepted definition of OO programming, it's more like an a la carte menu of options, pick at least X. Pick a description and set of assertions about why it's good, and I'd be happy to beat up that. Pro OO/Anti OO has been played out many times, and I was saying I basically agree with the arguments I've seen on the Anti OO side. Do you just want to see me repeat the arguments for why OO doesn't help with reuse or teaching programming and so forth? |