Comments on: Strange Names for Strange Days http://www.linux-mag.com/id/4137/ Open Source, Open Standards Sat, 05 Oct 2013 13:48:18 +0000 hourly 1 http://wordpress.org/?v=3.1 By: yoof http://www.linux-mag.com/id/4137/#comment-4624 yoof Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/4137/#comment-4624 There was a time when we designed in FORTRAN, but then coded the tricky bits (that needed to be efficient) in assembler macros. I wonder if now we should design with something like LISP (with OO and implicit parallelization) but code the tricky bits in C? There was a time when we designed in FORTRAN, but then coded the tricky bits (that needed to be efficient) in assembler macros. I wonder if now we should design with something like LISP (with OO and implicit parallelization) but code the tricky bits in C?

]]> By: nimityssj http://www.linux-mag.com/id/4137/#comment-4625 nimityssj Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/4137/#comment-4625 I've been pushing parallel languages for a while, and it's good to see a piece on it. I'm surprised the author didn't mention MIT's Cilk, which adds like 3 statements to the C lang. and the runtime parallelizes from there. The Nesl language, from CMU, has been used on numerous clusters, and supposedly can reduce code size significantly: on their quicksort it was "10 lines of code vs. 1700", compared to C++/MPI. Executed effeciently, too. I believe that if we are to deal with the complexity of parallel hardware, we will need better tools, like these, to do it. I’ve been pushing parallel languages for a while, and it’s good to see a piece on it. I’m surprised the author didn’t mention MIT’s Cilk, which adds like 3 statements to the C lang. and the runtime parallelizes from there. The Nesl language, from CMU, has been used on numerous clusters, and supposedly can reduce code size significantly: on their quicksort it was “10 lines of code vs. 1700″, compared to C++/MPI. Executed effeciently, too. I believe that if we are to deal with the complexity of parallel hardware, we will need better tools, like these, to do it.

]]> By: nimityssj http://www.linux-mag.com/id/4137/#comment-4626 nimityssj Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/4137/#comment-4626 TO YOOF:<br /> <br /> That was an excellent idea, and I have been asking the Erlang and Nesl people about this. I figure that if the critical, serial routines could be coded in C/C++ (i prefer C++), then incorporated into the parallel lang. code like how Java binds native code, then we could have the best of both worlds. Well, close enough.<br /> <br /> Is there anyone out there who develops with these languages or dev.'s the lang.'s themselves? Would like your input on this, but with Erlang (my favorite for HA systems), there are some limitations to this approach. Perhaps it will be addressed later. Might be better off just modifying a high-performance Common LISP system, like AllegroCL. HPC LISP (*LISP) was one of the first languages for this stuff, used on the Thinking Machines Corp.'s 64,000 processor systems. TO YOOF:

That was an excellent idea, and I have been asking the Erlang and Nesl people about this. I figure that if the critical, serial routines could be coded in C/C++ (i prefer C++), then incorporated into the parallel lang. code like how Java binds native code, then we could have the best of both worlds. Well, close enough.

Is there anyone out there who develops with these languages or dev.’s the lang.’s themselves? Would like your input on this, but with Erlang (my favorite for HA systems), there are some limitations to this approach. Perhaps it will be addressed later. Might be better off just modifying a high-performance Common LISP system, like AllegroCL. HPC LISP (*LISP) was one of the first languages for this stuff, used on the Thinking Machines Corp.’s 64,000 processor systems.

]]>