Slashdot Mirror
New Languages Vs. Old For Parallel Programming
joabj writes "Getting the most from multicore processors is becoming an increasingly difficult task for programmers. DARPA has commissioned a number of new programming languages, notably X10 and Chapel, written especially for developing programs that can be run across multiple processors, though others see them as too much of a departure to ever gain widespread usage among coders."
A lot of problems are I/O driven -- I would like to see more database client libraries allow a full async approach that lets us not block the threads we are trying to do concurrent work on.
Most computing intensive problems that a user will encounter at home are easily parallelizable, i.e. video encoding, gaming, photoshop filters, webbrowsing and so on. The amount of times where I maxed out a single CPU and the given problem would not have been to some large degree parallelizable are close to zero.
The trouble is that they are only "easy" parallelizable in concept, implementing parallelization on an exciting serial codebase is where it gets messy.
And how many of those cores are above 2% utilization for 90% of the day? Parallelization on the desktop is a solution is search of a problem- we have in a single core dozens of times what the average user needs. My email, web browsing, word processor, etc aren't cpu limited. They're network limited, and after that they're user limited (a human can only read so many slashdot stories a minute). There's no point in anything other than servers having 4 or 8 cores. But if Intel doesn't fool people into thinking they need new computers their revenue will go down, so 16 core desktops next year it is.
I still have more fans than freaks. WTF is wrong with you people?
I have not read the article (par for the course here) but I think there is probably some confusion among the commenters regarding the difference between multi-threading programs and parallel algorithms. Database servers, asynchronous I/O, background tasks and web servers are all examples of multi-threaded applications, where each thread can run independently of every other thread with locks protecting access to shared objects. This is different from (and probably simpler than) parallel programs. Map-reduce is a great example of a parallel distributed algorithm, but it is only one parallel computing model: Multiple Instruction / Multiple Data (MIMD). Single Instruction / Multiple Data (SIMD) algorithms implemented on super-computers like Cray (more of a vector machine, but it's close enough to SIMD) and MasPar systems require different and far more complex algorithms. In addition, purpose-built supercomputers may have additional restrictions on their memory accesses, such as whether multiple CPUs can concurrently read or write from memory.
Of course, the Cray and Maspar systems are purpose-built machines, and, much like special-build processors have fallen in performance to general purpose CPUs, Cray and Maspar systems have fallen into disuse and virtual obscurity; therefore, one might argue that SIMD-type systems and their associated algorithms should be discounted. But, there is a large class of problems -- particularly sorting algorithms -- well suited to SIMD algorithms, so perhaps we shouldn't be so quick to dismiss them.
There is a book called An Introduction to Parallel Algorithms by Joseph JaJa (http://www.amazon.com/Introduction-Parallel-Algorithms-Joseph-JaJa/dp/0201548569) that shows some of the complexities of developing truly parallel algorithms.
(Disclaimer: I own a copy of that book but otherwise have no financial interests in it.)
Sorry but, IMO, the cell is a perfect example of how not to design a multicore processor. Heterogenous processors introduce nothing new to the table of solutions that was not already there. We had systems with CPUs and GPUs before the Cell (or Intel's Larrabee and AMD's Fusion) showed up. Everybody knows that they're a pain in the ass to program. Neither CUDA nor OpenCL nor Microsoft's much ballyhooed TBB (threaded building blocks) will change that fact.
My point is that one does not design a parallel processor and then come up with a programming model to exploit it. It should be the other way around. The programming model should come first. One should design a model that makes parallel programming easy and the resulting apps rock-solid. Only then, after you have perfected your model, should you even consider designing a processor to support the model.
IOW, everybody's doing it wrong, and by everybody, I mean the all big players in the multicore hardware/software industry: Intel, Microsoft, IBM, Sun-Oracle, AMD, ARM, Apple, FreeScale, etc. The major computer science centers who are getting a lot of research money form the industry are not helping either since they have to kowtow to the likes of Intel and AMD whose main interest is to safeguard their installed base and preserve continuity.
It makes no difference. When the pain becomes unbearable (it's all about money), it will suddenly dawn on everybody that what is needed is to break away from the past.