Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 ggr 10/10/85; site bentley.UUCP
Path: utzoo!watmath!clyde!cbosgd!ihnp4!bentley!kwh
From: kwh@bentley.UUCP (KW Heuer)
Newsgroups: net.lang.c
Subject: Re: Long Longs
Message-ID: <608@bentley.UUCP>
Date: Mon, 3-Mar-86 16:13:02 EST
Article-I.D.: bentley.608
Posted: Mon Mar  3 16:13:02 1986
Date-Received: Wed, 5-Mar-86 03:56:50 EST
References: <3408@umcp-cs.UUCP> <491@faron.UUCP>
Organization: AT&T Bell Laboratories, Liberty Corner
Lines: 37

In article <491@faron.UUCP> faron!bs (Bob Silverman) writes:
>[How do I calculate without intermediate overflow]
>a = (b*c)/d;
>a = (b*c) % d;
>... Many machines CAN multiply two full words together giving a double
>length product placed in two adjacent registers. 

Many others can't.  If the feature were built into the language,
what code should it generate for "long long" arithmetic on a more
limited machine?

>Does C++ have long longs?

C++ cannot do anything that C cannot, since it produces C output.

>Does anyone have a good solution?

My prefered method (to maximize portability) is to have library
functions
	int muldiv(int, int, int)
	int mulrem(int, int, int)
which are written in assembly language; for specific applications
in tight loops I hand-optimize the .s file as described in article
<3408@umcp-cs.UUCP> by umcp-cs!chris (Chris Torek) (not quoted here).

It seems to me that best solution for the future is for "int" to be
32 bits wide, while "long" is 64.  I expect the compilers will start
doing this once there is a little more support for 64-bit arithmetic
in the hardware.  (Like when 32-bit addresses aren't big enough.)
No smiley on this paragraph, I'm serious.

Incidentally, does anyone have a muldiv routine (in C or as) for a
32-bit machine without emul/ediv instructions?  Currently I'm using
(int)((double)x*(double)y/(double)z) which is quite slow on a 3b2
without floating-point hardware.

Karl W. Z. Heuer (ihnp4!bentley!kwh), The Walking Lint