Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!husc6!cmcl2!rutgers!sri-spam!ames!ucbcad!ucbvax!ENGVAX.SCG.HAC.COM!KVC
From: KVC@ENGVAX.SCG.HAC.COM (Kevin Carosso)
Newsgroups: comp.os.vms
Subject: VAX C I/O performance
Message-ID: <8706260653.AA06090@ucbvax.Berkeley.EDU>
Date: Wed, 24-Jun-87 20:08:00 EDT
Article-I.D.: ucbvax.8706260653.AA06090
Posted: Wed Jun 24 20:08:00 1987
Date-Received: Sat, 27-Jun-87 07:09:30 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 27

I'm not much of a C programmer, so be kind... (but I'm an INFO-VAX old-timer!)

I've got a user with a C program that basically does read () and write ()
to move data from one file to another.  He says that it is terribly slow,
like one minute for a 1000 block data file.  Now, do most people find read
and write in C to be slow compared with (say) RMS calls?  or with read and
write on a suitably similar UNIX machine?  is there a big performance
difference between stream files and carriage return text files?

He says DEC told him that it's slow because write () won't write more than
512 bytes at a time.  Even if that's the case, should it be so slow?
My guess is that it only reads one records-worth and hence his code only
writes one records-worth.  Maybe it's that slow record-at-a-time.  If I'm
right, would the same apply to stream files?  (he says the input file was
created with a C program, so it could be a stream file).

Finally, should I just tell the guy that he has to use RMS or block-mode
QIO's for speed (that's what I'd do, but I'm not a C programmer).  He really
doesn't want to use any "horrible VMS system calls".

Anyway, I know the question is not posed very well, but if I get enough
input I can figure out what's going on and the best remedy.

Thanks,

        /Kevin Carosso                     kvc@engvax.scg.hac.com
         Hughes Aircraft Co.               kvc%engvax@oberon.usc.edu