kyle@xanth.UUCP (12/05/87)
[ Since this is not an EMACS issue I have redirected followups to comp.unix.questions. ] In article <3720010@hpsemc.UUCP>, bd@hpsemc.UUCP (bob desinger) writes: > Here's how it is on our HP-UX system, a model 840: > > drwxrwxr-x 2 bin mail 1024 Nov 25 18:45 /usr/mail > -rwxr-sr-x 2 root mail 137216 Oct 2 00:00 /bin/rmail Wow! Why is rmail so BIG? What does HP-UX rmail do that SMAIL 2.5 doesn't? Contrast the size of this rmail with various executables found on our 4.3 BSD system. -rwxr-xr-x 2 root staff 35840 Nov 3 07:02 /bin/rmail (SMAIL 2.5) -rwxr-xr-x 1 root staff 104448 Jun 5 1986 /lib/ccom (C compiler) -rwxr-xr-x 1 root staff 97280 Dec 5 05:17 /usr/local/carmen (Lisp) -rwsr-xr-x 1 root staff 100352 Apr 5 1987 /usr/lib/sendmail Just curious, kyle jones <kyle@odu.edu> old dominion university, norfolk, va usa
gwc@root.co.uk (Geoff Clare) (12/10/87)
>In article <3720010@hpsemc.UUCP>, bd@hpsemc.UUCP (bob desinger) writes: >> Here's how it is on our HP-UX system, a model 840: >> drwxrwxr-x 2 bin mail 1024 Nov 25 18:45 /usr/mail >> -rwxr-sr-x 2 root mail 137216 Oct 2 00:00 /bin/rmail >Wow! Why is rmail so BIG? What does HP-UX rmail do that SMAIL 2.5 >doesn't? Contrast the size of this rmail with various executables >found on our 4.3 BSD system. >-rwxr-xr-x 2 root staff 35840 Nov 3 07:02 /bin/rmail (SMAIL 2.5) >-rwxr-xr-x 1 root staff 104448 Jun 5 1986 /lib/ccom (C compiler) >-rwxr-xr-x 1 root staff 97280 Dec 5 05:17 /usr/local/carmen (Lisp) >-rwsr-xr-x 1 root staff 100352 Apr 5 1987 /usr/lib/sendmail The HP840 is a RISC architecture machine. Reduced instruction set implies more instructions required to do the same job than on a 'complex' instruction set machine, hence the proportionately larger executable files. Presumably your 4.3BSD machine is a VAX-alike (i.e. complex instruction set). The only other file from your list which exists on our HP840 system is the C compiler, and look at the size of that beast!! -rwxrwxr-x 1 bin bin 1097728 Mar 5 1987 /lib/ccom (No, that's not a typo - it really is more than 1 Megabyte!) Geoff Clare gwc@root.co.uk seismo!mcvax!ukc!root44!gwc -- Geoff Clare gwc@root.co.uk seismo!mcvax!ukc!root44!gwc
frank@zen.UUCP (Frank Wales) (12/15/87)
Be careful with these comparisons of size... In article <495@root44.co.uk> gwc@root44.UUCP (Geoff Clare) writes: >In article <3631@xanth.cs.odu.edu> kyle@xanth.cs.odu.edu (Kyle Jones) writes: >>In article <3720010@hpsemc.UUCP>, bd@hpsemc.UUCP (bob desinger) writes: >>> Here's how it is on our HP-UX system, a model 840: >>> -rwxr-sr-x 2 root mail 137216 Oct 2 00:00 /bin/rmail Well, I'm writing this on an HP 9000 model 840, and here is the result of ll-ing our /bin/rmail: -r-xr-s--x 2 root mail 71680 Mar 5 1987 /bin/rmail Maybe Bob's machine has a later release of HP-UX than ours, or maybe (being an HP machine) he's got a debuggable object, but nevertheless, our rmail works fine at only half that size. >>Wow! Why is rmail so BIG? What does HP-UX rmail do that SMAIL 2.5 >>doesn't? Contrast the size of this rmail with various executables >>found on our 4.3 BSD system. > >>-rwxr-xr-x 2 root staff 35840 Nov 3 07:02 /bin/rmail (SMAIL 2.5) >>-rwxr-xr-x 1 root staff 104448 Jun 5 1986 /lib/ccom (C compiler) >>-rwsr-xr-x 1 root staff 100352 Apr 5 1987 /usr/lib/sendmail > >The HP840 is a RISC architecture machine. Reduced instruction set implies >more instructions required to do the same job than on a 'complex' >instruction set machine, hence the proportionately larger executable files. >Presumably your 4.3BSD machine is a VAX-alike (i.e. complex instruction set). > >The only other file from your list which exists on our HP840 system is >the C compiler, and look at the size of that beast!! > >-rwxrwxr-x 1 bin bin 1097728 Mar 5 1987 /lib/ccom Well, we have a version of sendmail here too, and it's: -r-sr-s--x 3 root mail 147456 Sep 20 20:51 /usr/lib/sendmail Only half as big again as the CISC version, and indeed, that seems about average across most of the applications which we've moved from our older machines (HP 9000 Series 500s, proprietary complex 32-bit CPUs; sorry to see them go...). But wait, what about that enormous compiler? Well, here's our Pascal and Fortran compiler sizes too: -r-xr-x--x 1 bin bin 1718068 Aug 29 11:17 /usr/lib/f77comp -r-xr-x--x 1 bin bin 2445312 Aug 29 11:16 /usr/lib/pascomp Big, huh? Why is this? Because, as noted by Geoff, the HP 800 series are RISC-based machines, and one essential component of an effective RISC architecture is efficient optimising compilers; PCC and its cronies just won't pass muster. So HP rewrote all the compilers from scratch, and they are *much* better than any others I've seen on a Unix system. For example, cc *never* says: "syntax error"; you actually get told what the problem is, such as: "Subscript expression must combine a pointer and an integer". Note also that each compiler has a built-in optimiser (no separate /lib/c2, for example), and the FORTRAN and Pascal compilers emit object modules directly. But getting to the point -- the problem is simple: we're trying to compare across series and architectures only on file sizes, without considering whether the programs we're comparing actually do the same job in the same way; a bit like comparing motor cars purely on their length. If such size comparison is to have any kind of meaning (and I don't actually think it can, but I enjoy a good argument anyway), we need to compare programs whose precise behaviour is known to be the same. Even then, there's no guarantee that a particular application won't run very badly on a given architecture, or that a particular compiler won't optimise for speed in preference to size. Anyway, with this in mind, here's the size of a few well-known programs on our 840; cut and compare: -r-xr-x--x 1 bin bin 28672 Mar 5 1987 /bin/cat -r-xr-x--x 2 bin bin 77824 Mar 5 1987 /bin/ed -rwx--x--- 1 games play 423936 Dec 11 12:19 /usr/games/bin/nethack -r-xr-s--x 1 bin mail 264192 Nov 20 16:07 /usr/local/bin/elm -r-xr-x--x 1 bin bin 837632 Sep 15 09:45 /usr/local/bin/tex Frank Wales, Pointless Comparisons Engineer, [frank@zen.co.uk<->mcvax!zen!frank] Zengrange Ltd., Greenfield Rd., Leeds, ENGLAND, LS9 8DB. (+44) 532 489048 x220