Comments on: An Interview with Dr. Jay Hoeflinger about Automatic Parallelization http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/ A Blog on Parallel Programming and Concurrency by Michael Suess Sat, 31 Jul 2010 08:29:21 +0000 http://wordpress.org/?v=2.7 hourly 1 By: Deborah J. Meade http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/comment-page-1/#comment-97448 Deborah J. Meade Wed, 04 Feb 2009 13:24:32 +0000 http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/#comment-97448 hi and it is great to see how you are doing from a fellow DHS band member..........deborah j. meade hi and it is great to see how you are doing from a fellow DHS band member……….deborah j. meade

]]> By: Joseph Huang http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/comment-page-1/#comment-19666 Joseph Huang Thu, 16 Aug 2007 02:54:21 +0000 http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/#comment-19666 Thank you for removing the italics. <blockquote>There is a tension between the expressibility of a language and the ability of compilers to properly optimize it. A language with a high degree of expressibility may be easier for humans to use, yet at the same time it almost certainly is more complex and therefore more difficult for a compiler to understand, making it harder to parallelize.</blockquote> I think this only applies to impure languages. If you know order dependencies due to uniqueness typing (Clean, Mercury) or monads (Haskell) then a higher level language is easier to optimize because the optimizer knows which rearrangements are safe. Mercury is a general-purpose programming language. By the way, I didn't read the article at all before my previous comment. Thank you for removing the italics.

There is a tension between the expressibility of a language and the ability of compilers to properly optimize it. A language with a high degree of expressibility may be easier for humans to use, yet at the same time it almost certainly is more complex and therefore more difficult for a compiler to understand, making it harder to parallelize.

I think this only applies to impure languages. If you know order dependencies due to uniqueness typing (Clean, Mercury) or monads (Haskell) then a higher level language is easier to optimize because the optimizer knows which rearrangements are safe.

Mercury is a general-purpose programming language.

By the way, I didn’t read the article at all before my previous comment.

]]> By: Michael Suess http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/comment-page-1/#comment-19605 Michael Suess Wed, 15 Aug 2007 08:35:24 +0000 http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/#comment-19605 @Joseph: I will try to do something with regards to the readability of the text. With regards to your other comment: I know that there are languages that make automatic parallelization possible. In fact, there is even a question about it in the text. I am sure you will agree, however, that the language you mention (Mercury) is not exactly mainstream or in wide use... @Joseph: I will try to do something with regards to the readability of the text.

With regards to your other comment: I know that there are languages that make automatic parallelization possible. In fact, there is even a question about it in the text. I am sure you will agree, however, that the language you mention (Mercury) is not exactly mainstream or in wide use…

]]> By: Jan Liband http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/comment-page-1/#comment-19549 Jan Liband Tue, 14 Aug 2007 18:07:59 +0000 http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/#comment-19549 Dr. Hoeflinger notes that one of the most promising areas for automatic parallelization is in the area of domain-specific languages. While he cites examples of Fortran77 and Matlab, he did not mention SQL. SQL, and procedural forms of SQL such as DeviceSQL, provide the ease of programming of a functional, declarative language, while a SQL to C compiler can know a great deal about the internal data structures. A company called Encirq (www.encirq.com) has been working on such an approach for automatic parallelization. Applications include high-speed data stream and complex event processing, real-time analytics, network intrusion detection and more. A key advantage of this approach is the widespread expertise in SQL programming and the ease of integrating C code with existing applications. Dr. Hoeflinger notes that one of the most promising areas for automatic parallelization is in the area of domain-specific languages. While he cites examples of Fortran77 and Matlab, he did not mention SQL. SQL, and procedural forms of SQL such as DeviceSQL, provide the ease of programming of a functional, declarative language, while a SQL to C compiler can know a great deal about the internal data structures. A company called Encirq (www.encirq.com) has been working on such an approach for automatic parallelization. Applications include high-speed data stream and complex event processing, real-time analytics, network intrusion detection and more. A key advantage of this approach is the widespread expertise in SQL programming and the ease of integrating C code with existing applications.

]]> By: Enables new spec for enabling real time optimization of apps | insideHPC http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/comment-page-1/#comment-19535 Enables new spec for enabling real time optimization of apps | insideHPC Tue, 14 Aug 2007 15:47:17 +0000 http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/#comment-19535 [...] know enough about to parallelize at compile time in order to get the best performance (read yesterday’s post on Michael Suess’ blog for more on this last [...] [...] know enough about to parallelize at compile time in order to get the best performance (read yesterday’s post on Michael Suess’ blog for more on this last [...]

]]> By: Joseph Huang http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/comment-page-1/#comment-19527 Joseph Huang Tue, 14 Aug 2007 14:58:52 +0000 http://www.thinkingparallel.com/2007/08/14/an-interview-with-dr-jay-hoeflinger-about-automatic-parallelization/#comment-19527 Your article is impossible to read due to the italicized text. Automatic parallelization is already here. http://www.cs.mu.oz.au/research/mercury/information/papers.html#wangp-hons Your article is impossible to read due to the italicized text.

Automatic parallelization is already here. http://www.cs.mu.oz.au/research/mercury/information/papers.html#wangp-hons

]]>