Path: utzoo!attcan!uunet!cs.utexas.edu!sun-barr!sun!gammara!khb From: khb@gammara.Sun.COM (gammara) Newsgroups: comp.arch Subject: Re: Double Width Integer Multiplication and Division Message-ID: <113912@sun.Eng.Sun.COM> Date: 6 Jul 89 04:28:57 GMT References: <1046@aber-cs.UUCP> <1380@l.cc.purdue.edu> <13943@haddock.ima.isc.com> Sender: news@sun.Eng.Sun.COM Reply-To: khb@sun.UUCP (gammara) Organization: Sun Microsystems, Mountain View Lines: 42 In article <13943@haddock.ima.isc.com> suitti@haddock.ima.isc.com (Stephen Uitti) writes: >In article <1380@l.cc.purdue.edu> cik@l.cc.purdue.edu (Herman Rubin) writes: >(and various others wrote:) >very long). Compilers do not code as well as a trained human can >(this will probably always be true - on the other hand, compilers Depends on the compiler and the machine. The famed PL/.8 for the IBM 801 often produced provably perfect code (say those involved with the project). Try hand coding a Multiflow Trace/28 or a Cydra-5 ... from scratch (taking the f77 compiler output and hacking is cheating! :>). There are some humans who can do better than the compiler on these machines ... but for non-trivial codes the number of people is very small, and cost very high. The overall cost of a project is typically "maintence" bound (meaning changing the definition of the problem now that we have a trial solution ... all in the guise of making a few modest code changes). Then comes coding and design costs, etc. Handcrafted assembly language is seldom economically viable today. Given advances in compilers, and hardware evolution, it is often counter productive (coded SAXPY is slower than a compiler which inlines and outer loop unrolls). Specialized instructions which are of great interest to (say) 5% of the population are simply not worth putting into hardware (since it means that "real estate" can be dedicated to solving 95% of the worlds problems better). When folks have an instruction which is important enough to them (say $200M) they can probably get it in. The vast majority of purchasers have been voting with their feet ... simpler is more popular. (I am counting the 80xxx as simpler than a mainframe ... this is not just RISC vs. CISC this is thinking in economic, bottom line terms). Keith H. Bierman |*My thoughts are my own. Only my work belongs to Sun* It's Not My Fault | Marketing Technical Specialist ! kbierman@sun.com I Voted for Bill & | Languages and Performance Tools. Opus (* strange as it may seem, I do more engineering now *)