Sunday, May 27, 2007

Some have given us 50 years or so for things to last as we know them. What next, Planet of the Apes?? Back to stone age?? This short sighted world does not quite feel like home at all, i.e. days when I used to imagine how everything circled around compassion.

Coming to computer architecture, one the one hand you have David Patterson delivering a lecture in PARC forum on how it is going to be grand many core era and the inevitable slow death of things we hold near and dear, Pentium IV, AMD Opteron, Power4, Itanium, and not to miss the unscalable yester years' misguiding notions like Alpha EV8 or 21464, notions @ On the other hand you have venture capitalists claiming that the investments in Silicon are pointless. They are looking to sponsor projects only in Web 2.0! Unfortunately the next big thing will be Web 2.0, it seems. This Web 2.0 is all hype, right. Most of the smart audio projects seem to be circling around ipods. As one who does not possess ipod, I find this rather pointless. Secondly all this cool apps like, google word/excel etc web apps do not hold water. And they are not like a technical marvel. There is DocBook and likewise XML driven apps. So a tree structure intermediate language i.e. XML is primarily the basis for everything? But thinking positively does it mean revival of great stuff like Latex, which people never really adapted in a broad scale? Does not make much of a difference to me, especially and to some others I know. I have used Latex for a while and it has been my tool of choice to write a technical document.

Of late I am taking interest in data flow exectution machines. I am working on the compiler for a processor which has no register files or a common store as we know them. i.e it does not have a notion of a stack for parameter passing. What the is the computation model you may ask. Values do not stay at a place for more than a clock tick or two. Generated data must be either consumed or passed around to be used later. Complications arise especially if you try to compile to this from an imperative language like C, which is what they (including me) are trying to do here over the years... but there is just so much to do.

I am exploring if Dr. Kesav Pingali's work: "Dependency flow graphs: An algebraic approach to Program Dependencies" apply here. More on it later!

No comments: