Comments on: What Makes Parallel Programming Hard? http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/ A Blog on Parallel Programming and Concurrency by Michael Suess Sat, 31 Jul 2010 08:49:34 +0000 http://wordpress.org/?v=2.7 hourly 1 By: Making Bank on DART » Blog Archive » What Makes Parallel Programming So Difficult? http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/comment-page-1/#comment-19640 Making Bank on DART » Blog Archive » What Makes Parallel Programming So Difficult? Wed, 15 Aug 2007 20:14:46 +0000 http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/#comment-19640 [...] This blog post from Intel clearly lays out the challenges of parallel-programming, and this response to the post from Thinking Parallel points out even more complications of parallel models of programming. The [...] [...] This blog post from Intel clearly lays out the challenges of parallel-programming, and this response to the post from Thinking Parallel points out even more complications of parallel models of programming. The [...]

]]> By: Nick Vaidyanathan http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/comment-page-1/#comment-19134 Nick Vaidyanathan Wed, 08 Aug 2007 21:28:13 +0000 http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/#comment-19134 Nice article. I'm curious if you've studied the differences/similarities between concurrent programming and hardware design. I did some VHDL in undergrad and had to deal with the paradigm shift of dealing with processes running in parallel, variables versus signals, etc...it really helped when I was learning about threads in OS (which I've admittedly never had to use in an enterprise app), and I'm wondering if there are methodologies/techniques in that world could bridge the gap and help in software. Nice article. I’m curious if you’ve studied the differences/similarities between concurrent programming and hardware design. I did some VHDL in undergrad and had to deal with the paradigm shift of dealing with processes running in parallel, variables versus signals, etc…it really helped when I was learning about threads in OS (which I’ve admittedly never had to use in an enterprise app), and I’m wondering if there are methodologies/techniques in that world could bridge the gap and help in software.

]]> By: Fernando http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/comment-page-1/#comment-19071 Fernando Tue, 07 Aug 2007 22:13:20 +0000 http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/#comment-19071 I saw great benefits in converting a parser algorithm from sequential to parallel. Basically i figure that for parallel to work correctly the tasks that you want to execute in parallel must not be dependant on each other. In Java I heavily relied on the Concurrent framework. The algorithm i tested scales very well compare to its sequential counterpart. I saw great benefits in converting a parser algorithm from sequential to parallel.
Basically i figure that for parallel to work correctly the tasks that you want to execute in parallel must not be dependant on each other.
In Java I heavily relied on the Concurrent framework.
The algorithm i tested scales very well compare to its sequential counterpart.

]]> By: Seun Osewa http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/comment-page-1/#comment-18953 Seun Osewa Mon, 06 Aug 2007 19:46:23 +0000 http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/#comment-18953 My guess is that most applications that require parallel programming to run well on current hardware are extremely complex problems. Complex enough that the overhead of learning parallelism is just an extra hurdle to scale. In that case, current languages already offer sufficient support for any style of concurrency you happen to prefer (CSP with processes and pipes, SMT with threads, and event-based programming with libraries like libevent). It's not a big deal. My guess is that most applications that require parallel programming to run well on current hardware are extremely complex problems. Complex enough that the overhead of learning parallelism is just an extra hurdle to scale. In that case, current languages already offer sufficient support for any style of concurrency you happen to prefer (CSP with processes and pipes, SMT with threads, and event-based programming with libraries like libevent). It’s not a big deal.

]]> By: Minesh B. Amin http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/comment-page-1/#comment-18943 Minesh B. Amin Mon, 06 Aug 2007 18:07:22 +0000 http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/#comment-18943 Hi Michael, Your list is right on the mark. The thing that I find the most fascinating is the lack of a methodology or a library that bridges the gap between the goal (scalable solutions) and the reality (existing serial software processes/developers). Well not quite ... If interested, please come-by and try out the open source version of my company's solution (OpenSPM). Very soon, I will be going live with a collection of parallel dynamic languages (starting with parallel TCL). Let me know if you would like to try out parallel TCL before general availability. Best regards. Hi Michael,

Your list is right on the mark. The thing that I find the most fascinating
is the lack of a methodology or a library that bridges the gap between
the goal (scalable solutions) and the reality (existing serial software
processes/developers). Well not quite …

If interested, please come-by and try out the open source version
of my company’s solution (OpenSPM). Very soon, I will be going
live with a collection of parallel dynamic languages (starting with
parallel TCL). Let me know if you would like to try out parallel TCL
before general availability.

Best regards.

]]> By: anon http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/comment-page-1/#comment-18925 anon Mon, 06 Aug 2007 14:33:49 +0000 http://www.thinkingparallel.com/2007/08/06/what-makes-parallel-programming-hard/#comment-18925 Add another: It's practically impossible to know when you can compose parallel software libraries. You are stuck with either reimplementing / reengineering large chunks of code or limiting yourself to partial parallelism by separating the code into disjoint services. Often those services wait on each other, so you're just adding overhead and losing a portion of you machine or allocation. Add another: It’s practically impossible to know when you can compose parallel software libraries. You are stuck with either reimplementing / reengineering large chunks of code or limiting yourself to partial parallelism by separating the code into disjoint services. Often those services wait on each other, so you’re just adding overhead and losing a portion of you machine or allocation.

]]>