PM, On Sat, Oct 18, 2014 at 6:24 PM, Piaget Modeler via AGI <[email protected]> wrote:
> Languages like Fortran are programming to the computing device. > FORTRAN and COBOL were two of the LEAST machine-specific languages around. Kernigan and Ritchie even explain how C was optimized for PDP-11 computers. FORTRAN efficiently made the jumps from vacuum tube computers all the way to full vector supercomputers and remains the "language to beat" on supercomputers. Ben had it a bit wrong regarding legacy programs. The truth of the matter is that "big iron" programs like weather prediction have been around for a while, and they would do ANYTHING to them if they thought they could squeeze a little more speed from them. Alittle more speed maps into another day in advance that they can predict the weather. They stick with FORTRAN because it remains the fastest language around. But, how can this be when you can do SO much more in C/C++? The simple answer is that NONE of the bells and whistles in C make computing any more efficient. However, they all get in the way of efficient vectorization. > > I submit to you that we need to devise a language of mental operations > that an AGI will need to execute, rather than machine operations. > That is one theory subscribed to by many on this forum. However, there aren't many others in the world who believe this. As I have explained in the past, our views of what we do are but mental models that in all probability have absolutely NOTHING to do with how we actually operate. This is an illusion that people on this forum believe, with all of the fervor and lack of evidence of any other religious zealot. You would be MUCH better off by NOT building such stuff into the language, but rather providing a mechanism to extend the language in whatever manner future research leads - including mental operations if that is your pleasure. And, since you will be programming in your own extensions, the basic language should be "down and dirty" simplistic to facilitate maximum efficiency. So, we are back to FORTRAN and COBOL. Steve > > ------------------------------ > Date: Sat, 18 Oct 2014 13:24:14 -0700 > Subject: Re: [agi] Event Models > From: [email protected] > To: [email protected] > > John, > > On Sat, Oct 18, 2014 at 6:33 AM, John Rose via AGI <[email protected]> > wrote: > > Here's an idea - let's re-spec Fortran to do better 3D graphics rendering > and real-time physics simulation so people can use it to write video games. > The question is why? Perhaps to migrate existing developers? Better to > create a whole new language IMO. > > > OK, on the way to answering this question, I'll ask and answer the > following question: > > Q: Why do most supercomputers STILL program in FORTRAN?!!! > > A1: Because it has carefully conceived language impediments that make it > much easier (than C, Java, etc.) to write code in ways that are > automatically vectorizeable. This started for different reasons, e.g. the > motivation to use the TIX (conditional Transfer and Increment an indeX > register) instructions on the early IBM 70X computers, later to become 70XX > computers. However, the requirements to write code as efficient loops are > nearly the same in vector machines. > > A2: The complexity of an optimizing and vectorizing compiler tends to > grow as the square of the number of elements in the language, so compilers > for "modern" languages tend to do a bad job of optimization, and a horrible > job of vectorization, unless you write in a minimal subset of the languages > that strongly resembles FORTRAN. > > Of course if your problem is SO simple it can be run on an Intel > processor, which clearly does NOT include AGI, then there is no need for > executional efficiency. > > You mentioned rendering. Have you looked at the execution times for > rendering programs? These days, they must schedule planned movie releases > around the expected rendering times. Is it REALLY worth such costs to > implement more convenient language constructs? > > Language impediments were carefully studied when FORTRAN was created, but > this art has been lost to time. Perhaps you remember the early restriction > that subscripts had to be of the form aX+b? This restriction was soon > removed, but expert programmers learned to abide by the removed > restrictions, because doing so made their code run faster. > > In implementing something like COBOL, the issue of language restrictions > should be carefully revisited. > > Steve > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/10443978-6f4c28ac > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com > > > > > -- > Full employment can be had with the stoke of a pen. Simply institute a six > hour workday. That will easily create enough new jobs to bring back full > employment. > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/19999924-4a978ccc> | > Modify <https://www.listbox.com/member/?&;> Your Subscription > <http://www.listbox.com> > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/10443978-6f4c28ac> | > Modify > <https://www.listbox.com/member/?&;> > Your Subscription <http://www.listbox.com> > -- Full employment can be had with the stoke of a pen. Simply institute a six hour workday. That will easily create enough new jobs to bring back full employment. ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
