home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.benchmarks:1878 comp.arch:11757 comp.misc:4636
- Newsgroups: comp.benchmarks,comp.arch,comp.misc
- Path: sparky!uunet!rde!ksmith!keith
- From: keith@ksmith.uucp (Keith Smith)
- Subject: Re: Who wants faster machines?
- Organization: Keith's Computer, Hope Mills, NC
- Date: Thu, 17 Dec 92 03:59:31 GMT
- Message-ID: <1992Dec17.035931.1180@ksmith.uucp>
- Followup-To: comp.misc
- References: <1f2qefINNfaq@uranium.sto.pdb.sni.de> <1992Dec4.103129.524@fasttech.com>
- Lines: 54
-
- In article <1992Dec4.103129.524@fasttech.com> zeke@fasttech.com (Bohdan Tashchuk) writes:
- >In <1f2qefINNfaq@uranium.sto.pdb.sni.de> mollers.pad@sni.de (Josef Moellers) writes:
- >
- >>I'd say that 60% of all added performance is easily
- >>souped up by improvements in the man-machine interaction.
- >
- >>THAT's what we need fast machines for!
- >
- >About 20% is useful improvement and about 80% counteracts software bloat.
- >
- >The architects and h/w guys keep making machines faster and faster and the
- >software keeps getting less and less efficient.
-
- Sofware bloat, software bloat, naa, naa, 'naa, naa, ..., naa.
-
- Jeez.
-
- I look at some of the code I wrote for my Z80 in assembler. Zillions of
- hours, for not much of a program. Spaghetti code to the max.
- Commented, and I STILL can't follow what I was doing. With the current
- cost of cheap storage, and CPU cycles, I'll gladly take a little
- software bloat, for simple readability, and maintainability, along with
- avoiding alot of hand coding in assembler for speed & size.
-
- I'll also disagree with the 20/80 ratio you have come up with, as it
- relates to performance. What we are seeing today in the business world
- are *STABLE* *RELIABLE* database systems that run on commodity hardware,
- in a flexable enviroment. This same functionality 15 years ago would
- have required an IBM 34/36 style equivilance, with comperable amounts of
- fixed storage. Indeed, I replaced same with a 33Mhz 32 bit "pee cee", as
- opposed to a 16Mhz 32 bit Midrange. The performance was *FAR* in excess
- of doubled, and yet the storage requirements, and program sizes actually
- SHRUNK.
-
- As the software that runs more complicated systems get's more
- complicated (multi-user, multi-task, journaling filesystems, graphical
- interfaces, real-time processing ...) the programs *HAVE* to increase in
- complexity at a similar expotential rate. The number of interactions
- between different software modules will increase expotentially also, as
- you add on new modules to handle ever more complex tasks, as each task
- that gets effected will increase in size.
-
- Writing spaghetti code that weaves thru 50 sections of code blind fast
- is great, and you can probably follow it. Writing same thru 1000
- sections of code is simply idiotic to consider. In turn the code in
- each section will swell, as it handles more function in order to
- insulate itself from a breakdown elsewhere, then splits, etc. Is it
- bloat? Yes. Is it bad? No, simply a neccessary part of software
- evolution IMHO. Does it affect performance? Sure! Now we can run the
- machines all day without reseting it 5 times ...
- --
- Keith Smith uunet!ksmith!keith 5719 Archer Rd.
- Digital Designs BBS 1-919-423-4216 Hope Mills, NC 28348-2201
- Somewhere in the Styx of North Carolina ...
-