5 Apr 2006

winterkoninkje: shadowcrane (clean) (Default)

There's a conceptual problem that's been plaguing me with Eng. For a first writing on the language, this isn't the area I would normally choose. But the most logical starting points — e.g. lexical aliasing, code metastructure — I've a pretty good handle on. Those are topics I will delve into in the future, but for now, this is my problem.

One of the design considerations for Eng is to be able to provide crossplatform natively parallel code. [1] When I say this, one should bear in mind that this goal alone is as large or larger than all the other goals for the language combined. There's a specific facet to a specific part of my attempt to do this which is causing me troubles, but before that, some history.

Back in the 1980s parallelism was very much in vogue. Then, come the late-1980s/early-1990s when microprocessor speeds started bounding through the roof, everyone decided parallelism was stupid and abandoned most research on it. (Though of course, there are always a few diehards for any technology.) These days, as we're starting to reach the limits of current fabrication technologies, and with the shape of the curve between speed and cost and how it's evolved over time, the field of parallelism is starting to simmer again with a passion fit to boil over the coming decade.

To see this, just take a look at the current market: basic home computers are shipping with 2~4 processors, many of those are dual- or (will soon be) quad-core processors, or have hyperthreading. And if that home market doesn't convince you, then take a look at console game platforms like the xbox360 or the ps3 (or their previous counterparts). And this doesn't even get into the server market which is growing ever faster as computers become persistantly more ubiquitous tools for every endeavor and every business. Nor does it get into distributed systems like the microchip in your keyboard and mouse. Let alone the scientific markets for weather modelling, VLSI, and AI — i.e. those diehards who never left.

...

Now that you're convinced that a natively-parallel language is essential, the first part of my problem.  )

June 2017

S M T W T F S
    123
45678910
11121314151617
18192021 222324
252627282930 

Tags

Page generated 26 Sep 2017 07:09 am
Powered by Dreamwidth Studios