Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site nrcvax.UUCP
Path: utzoo!linus!decvax!ittatc!dcdwest!sdcsvax!sdcrdcf!psivax!nrcvax!jj
From: jj@nrcvax.UUCP (Utah)
Newsgroups: net.lang.c
Subject: Re: Right shift vs. divide
Message-ID: <449@nrcvax.UUCP>
Date: Tue, 24-Dec-85 12:37:21 EST
Article-I.D.: nrcvax.449
Posted: Tue Dec 24 12:37:21 1985
Date-Received: Sat, 28-Dec-85 13:02:00 EST
References: <974@brl-tgr.ARPA>
Reply-To: jj@nrcvax.UUCP (Jeff Jennings - NRC-Utah)
Organization: Network Research Research Corp.  Oxnard,CA
Lines: 23

In article <974@brl-tgr.ARPA> Schauble@mit-multics.arpa (Paul Schauble) writes:
>
>BTW, is >> defined to be a sign-extended shift or a zero extended shift
>and under what circumstances.
>  
>                                    Paul
>                                    Schauble @ MIT-Multics

The way that I interpreted this when generating machine code (68000) for a >> X
was that if a was 'signed' (char, short, int, and long in our implementation)
the compiler generated an arithmetic shift.  If a was 'unsigned'
(unsigned char, unsigned short, unsigned int, unsigned long), it generated
a bitwise (logical?) shift.

BTW, most C compilers I have seen do not fold the following constant expression
correctly:	((unsigned long)0xFFFFFFFF) >> 1 /* or anything w/sign bit on */
Seems like the constant folding function in the compiler only does arithmetic
shifts.
-------------------------------------------------------------------------
Jeff Jennings		      		Network Research Corp.
ihnp4!nrcvax!nrcutah!jj			923 Executive Park Drive Suite C
ucbvax!calma!nrcvax!nrcutah!jj		Salt Lake City, Utah 84117, U.S.A.
{sdcsvax,hplabs}!sdcrdcf!psivax!nrcvax!nrcutah!jj	(801) 266-9194