Comments on: How The GPL Can Save Your Ass

By: cdsteinkuehler

cdsteinkuehler — Wed, 30 Nov -0001 00:00:00 +0000

Um…so what do you think about projects like Intel’s TBB: http://threadingbuildingblocks.org/

It’s GPL, and provides high-level abstraction for implementing parallelism. Or am I missing something in your requirements (perhaps you despise C++?!?) :)

By: wambamboozle

wambamboozle — Wed, 30 Nov -0001 00:00:00 +0000

I agree we should be using the GPL, especially for research.

Although I prefer declarative languages like Erlang for parallel programming, there is a widely deployed alternative.

Sun has been making multicore machines for many years now. The designers of J2EE provided a framework so that you can write multithreaded software without really thinking about it much. I've been on machines with hundreds of cores serving up webapps without a problem. A lot of this stuff has been worked out. You can't argue with working code.

By: pahosler

pahosler — Wed, 30 Nov -0001 00:00:00 +0000

This would have been an interesting read if not for the irritating spelling and punctuation errors! Please don’t give the “English is my second language” excuse, you’re a SENIOR EDITOR!!! If English is not your second language, please stop relying on MS Office to grammar check for you, it’s getting it wrong and making your article appear to be written by a 5th grader. Sorry, I really just expect more from folks that have Senior Editor as a title. Then again, I guess it doesn’t mean much anymore these days.

By: sjinsjca

sjinsjca — Wed, 30 Nov -0001 00:00:00 +0000

"...multi-core is parallel computing and parallel programming is hard, expensive, and in some cases non-portable..."

Aside from the GPL issues you raise... ever hear of LabVIEW? It's the leading programming environment for instrumentation. For more than twenty years it has made programming parallel processes easy, and it has done a fine job of simulating parallelism on the single-CPU architectures typical until recently. Then it was extended to programming FPGAs (field-programmable gate arrays, basically build-your-own-silicon devices), where a typical user can easily create many dozens of truly parallel processes running simultaneously. From there it was a small matter to support multicore processors, and that was folded into the current release of LabVIEW last year.

And LabVIEW runs on Windows, Mac OSX, Linux, Solaris... your code is entirely portable, and if you have more than one processor core, it'll take advantage of it.

It's marvelous, marvelous stuff. I really hope it represents the future of programming, because it puts tremendous power in the hands of ordinary folks like you and me. FPGAs are a good example-- before, it took special training, skills and costly tools (and lots and lots of engineering and test time) to design an FPGA. LabVIEW makes it easy, and turns Field Programmable Gate Arrays into USER-Programmable Gate Arrays. That's astonishing new functionality, and anyone who uses LabVIEW can handle it.

By: lry198010

lry198010 — Wed, 30 Nov -0001 00:00:00 +0000

so,How about the erlang! what’s the role of erlang!

By: quickening

quickening — Wed, 30 Nov -0001 00:00:00 +0000

Folks here mentioning programming solutions and I think that misses the big picture. Handling multiple cores should be no different than handling multiples of any other hardware resource – a task suited best for the OS. Multiple cores should be considered an opportunity to implement advances in computer intelligence such as the community of specialized agents concept.

By: bradlepc

bradlepc — Wed, 30 Nov -0001 00:00:00 +0000

The point of the article is that the solution needs to be broader and more easy to retrofit than what existing options offer. There’s billions of dollars worth of software out there that needs to stay on the performance curve, but for which “rewrite it in J2EE or Labview” is clearly not a viable answer. It would be great if the OS could just handle it, but the OS has little visibility into the inherent parallelism of an application. TBB is closer to the right idea, but it is still closely identified with Intel and its scope is relatively narrow at present.

By: junnufunky

junnufunky — Wed, 30 Nov -0001 00:00:00 +0000

I agree with the editor that (we) programmers are really facing a challenge in front of the parallel programming paradigm. And we really need some new tools and thinking how to use the resources the new architectures provide.
But as if GPL would be the answer to the problems? My personal experience is that if, for example, libraries and tools are licenced with LGPL-type licence, the following business model covers much more ground compared to being licenced by plain GPL. There are numerous examples of high quality tools and libraries produced this way with remarkable developement input from commercial users.
To consider that GPL is the silver bullet that solves the problem…not! Much more can be gained
if the interests of commercial and non-commercial communities can be unified. What results is useful for parties developing commercial and non-commercial software. Is there something wrong with this scenario?

By: hhemken

hhemken — Wed, 30 Nov -0001 00:00:00 +0000

No doubt this proposal will be derided as hare-brained commie pinko bullshit of no relevance to civilized capitalist society, but I think it is pretty much on the mark.

Parallelism needs to be a low-level functionality that is reused by developers as if it were just another library. Standardized use-cases should be supported, as well as the flexibility to design more specialized software mechanisms.

Even Aunt Tilly and Grandma will make full use of their multicore personal supercomputers once everyday software arrives making use of open-source parallelism libraries and techniques. Making videos, music, renderings of various kinds, and a variety of other heavy-duty simulation functionality will be subsumed into industrial, professional, and even consumer software.

By: nihil75

nihil75 — Wed, 30 Nov -0001 00:00:00 +0000

I think saying we need a solution before we are "knee deep in water" is an exaggeration.

The big corps don't mind waiting a few years,
They will profit alot more from having patented the solution as their own.

The only ones who are loosing right now are the end users, who cant utilize this technology.

By: pmpope

pmpope — Wed, 30 Nov -0001 00:00:00 +0000

Wow! I guess if I make as many coding mistakes as you do grammatically I could be some sort of senior and blame it on the fact that I’m a writer and not a coder. WOW! God bless Amerika!

By: grndrush

grndrush — Wed, 30 Nov -0001 00:00:00 +0000

I see nothing ‘new” or “innovative” in this article, but one thing you need as badly as top-notch parallel programmers, is a top-notch technical editor. Both the lack of substance AND the copy, so bad it hurts to read (if you’re a technical copy-editor, anyway), indicate you started writing this about 2 hours before deadline.

By: hacklinuxdude

hacklinuxdude — Wed, 30 Nov -0001 00:00:00 +0000

Microsoft does not like this and is trying to make GPL non free by patent threats.

http://www.digitalmajority.org/forum/t-49513/brad-smith-continues-its-fud-spreading-wants-to-tax-redhat

I dont know how long can GPL stand againts the $250 billion godzilla that microsoft is.

By: davrusso

davrusso — Wed, 30 Nov -0001 00:00:00 +0000

By and large it appears to me that the vast majority of users use their computers for nothing more than word processing, email, and accessing databases. At this point the average PC strikes me as satisfactory with a single-core CPU. Even my parametric solid-modeling CAD and small engineering simulations work well with a single CPU. The extra CPU’s are helpful though to parallel-process a CFD problem or whatever other science you’re into. Perhaps this should really be the direction of the industry, developing the parallel processing towards those who need it, using a generic MPI or something.