home *** CD-ROM | disk | FTP | other *** search
- Path: engr.sgi.com!shankar
- From: shankar@engr.sgi.com (Shankar Unni)
- Newsgroups: comp.std.c
- Subject: Re: [Q] 16bit, 32bit, 64bit ints, longs etc.
- Date: 22 Jan 1996 02:55:35 GMT
- Organization: Silicon Graphics, Inc., Mountain View, CA
- Distribution: world
- Message-ID: <4duub7$3cf@fido.asd.sgi.com>
- References: <4dtgug$5mf@cc.tut.fi>
- Reply-To: shankar@engr.sgi.com
- NNTP-Posting-Host: boris.engr.sgi.com
- X-Newsreader: TIN [version 1.2 PL2]
-
- Koivisto Hannu (k152608@proffa.cc.tut.fi) wrote:
-
- > 64bit. You see, I thought that in 64bit systems ints would be 64bit and
- > longs 64bit. However, when I tried this with gcc running on an Alpha
- > machine, it turned out that ints are 32bit and longs are 64bit(of course,
- > pointers are also 64bit).
-
- [ Non-standards-related digression ]
-
- This is usually a matter of trading off programming convenience against
- loss of performance.
-
- If every loop counter and temporary value (ints) were 64 bits, your memory
- requirements could balloon. On the other hand, having ints be 32 bits means
- that if you forget to declare malloc() (a common failing in old code), you
- could end up with garbage..
-
- It's a judgement call.
-
- Incidentally, many vendors are starting to ship files like "inttypes.h" (I
- know Solaris and Irix have this file; and I'm sure many others do as
- well), to declare integral types of known size, like "int32_t", "int64_t",
- etc., so that you can choose exactly what sizes you want regardless of the
- architecture you are porting your code to.
-
- --
- Shankar Unni E-Mail: shankar@sgi.com
- Silicon Graphics Inc. Phone: +1-415-933-2072
- URL: http://reality.sgi.com/employees/shankar
-