liam@cs.qmw.ac.uk (William Roberts) (04/06/90)
>So, my question is: If the Apple fortran is so much slower than the >NKR fortran, are the third party C compilers for A/UX better than the >Apple C compiler for A/UX??? The gcc compiler is a great deal better than the standard A/UX C compiler - it is also better than the standard Sun C compiler, the standard HP C compiler etc etc. It is also free, which is very good value for money! Dhrystones on the Mac IIfx cc, no register declarations 9,900 cc, with register declarations 11,100 gcc 12,500 Note that the IIfx has cache memory, which helps to overcome the feeble register allocation of the standard cc compiler. [[ Observant readers may know that the IIfx only runs A/UX 2.0: the figures quoted are for 1.1 binaries - yep, seems to be code compatible with 1.1 for straight A/UX system calls. Similar figures (albeit smaller) are obtained from the IIcx though the no register version is even worse still. ]] How about it Apple? I know that Stallman hates you, but if you stick to the letter of his "What you can and can't do" notice on gcc then you could distribute it with A/UX 2.0 and make lots of people very happy. I did try the Unisoft Pascal Compiler under A/UX 1.1, but it was not capable of handling real "Pascal for teaching computer science"; for example it generated illegal instructions for non-local gotos. If it's a validated ISO Pascal compiler then I can suggest a few extra tests for the validation suite...:-) -- William Roberts ARPA: liam@cs.qmw.ac.uk Queen Mary & Westfield College UUCP: liam@qmw-cs.UUCP Mile End Road AppleLink: UK0087 LONDON, E1 4NS, UK Tel: 01-975 5250 (Fax: 01-980 6533)
name@portia.Stanford.EDU (tony cooper) (04/08/90)
In article <1881@sequent.cs.qmw.ac.uk> liam@cs.qmw.ac.uk (William Roberts) writes: >The gcc compiler is a great deal better than the standard A/UX >C compiler - it is also better than the standard Sun C In addition to cc and gcc there is the Green Hills C compiler ported to A/UX by Unisoft. Here are dhrystone figures for the three on my machine: 7009 cc -O 8522 gcc -O (GNU) 9868 gcc (Green Hills) How do they compare for floating point calculations? Here are Whetstones (double precision): 1330 cc -O 2090 gcc -O (GNU) 2520 gcc (Green Hills) How about floating point performance using Linpack? For double precision using rolled BLAS we get in kflops: ---- cc -O (wouldn't compile, ran out of tree space) 259 gcc -O (GNU) 273 gcc (Green Hills) The Green Hills compiler does a good job of putting in inline 68881/2 floating point calls. the GNU compiler is faster when -fstrength-reduce is not used. How about Fortran compilers? There is a Green Hills f77 optimizing compiler, the A/UX 1.1 Fortran compiler and f2c which converts Fortran to C. Here are the double precision kflops for Linpack: 193 f77 -O 251 gf77 (Green Hills) To use f77 -O requires editing link assembly statements into link.l otherwise the optimizer chokes. Tony Cooper tony@popserver.stanford.edu