What to Watch for in 2007

An anonymous reader writes "InformationWeek picks its '5 Disruptive Technologies To Watch In 2007.' The list, which is based on the idea that these are areas which will move into the mainstream this year, includes RFID, graphics processing engines, server virtualization, Web services, and mobile security." What made your list?

Comments

Less clicking

[reset_button]

Entire article on one page

Multicore goes mainstream

[earthforce_1]

Massively parallel software development will move towards the mainstream as CPUs with 4 or more cores start to become mainstream. Inherently parallel languages such as system C intended for hardware design (and never really took off in this arena) may garner a second life as a way to reuse C/C++ libraries in environments with large numbers of processor cores running in parallel. Software engineers will eventually have to wrap their brain around the concepts found in HDL languages such as Verilog/VHDL whee everything is assumed to happen in parallel, with program state changes at defined synchronization/clock intervals.

Re:Multicore goes mainstream

[ClosedSource]

I wouldn't go too far with the hardware analogy. Synchronization in digital hardware is needed because of the unequal delays in different gate paths and determining the worst case timing is trivial by comparison to a similar calculation on software functions. On modern PC's caches can cause significant timing variations so synchronization based on time intervals would be quite problematical. Of course, one could probably choose a time interval so long that these variations would be swallowed, but you'd have to be willing to waste a lot of cycles which would defeat the purpose of adding multiple cores.

How about getting facts straight...

[ameline]

From the article; "Nvidia has had its CUDA program for several years now to assist developers that want to harness their graphics engines for computational applications."

Nuh uh. CUDA is new with the G80. They may have had something going, but it wasn't cuda.

As for being disruptive -- maybe using the GPU for computation will speed some things up -- those things that are extremely parallelizable, and single precision FP -- thats about it. The GPUs are not easy to program to -- CUDA is pretty tricky, and it's fairly well tied to nVidia's new architecture (I don't see ATI adopting it). The stuff from PeakStream and RapidMinds is a bit higher level, and can work on both ATI and nVidia chips, both have their pros and cons. It's early days yet for this -- I don't see it catching on in a big way for another couple of years. Then I think it will catch on in a big way -- but the tools are too immature at the moment for that to happen, and it's hard to predict what is going to catch on. Anyone interested in this stuff should be paying close attention to all of them -- I know I am.

See it all at once!

[Anonymous Coward]

Want to see the whole article at once? So do I!

[printable version]
http://www.informationweek.com/shared/printableArt icle.jhtml?articleID=196800208

Re:Low latency

[passthecrackpipe]

And what do you think most DSL runs on? Those engineers had a point.