Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!nic.MR.NET!xanth!mcnc!ncsuvx!cccvs1.ncsu.edu!rcb
From: rcb@cccvs1.ncsu.edu (Randy Buckland)
Newsgroups: comp.sys.mac
Subject: Re: New Mac Rumours
Message-ID: <2606@ncsuvx.ncsu.edu>
Date: 28 Feb 89 14:06:21 GMT
References: <IY0io5y00UgPAM2b4a@andrew.cmu.edu> <41a2364a.a590@mag.engin.umich.edu> <70755@ti-csl.csc.ti.com> <41aa5eaa.a590@mag.engin.umich.edu> <15682@versatc.UUCP>
Sender: news@ncsuvx.ncsu.edu
Reply-To: rcb@cccvs1.ncsu.edu.UUCP (Randy Buckland)
Organization: North Carolina State University
Lines: 31

For general amusement, try looking at the generated code for the following
in LightSpeed C 3.0

	strcpy (out, in)
	char *out, *in;
	{
		while (*in)
			*out++ = *in++;
	}

as opposed to

	strcpy (out, in)
	char *out, *in;
	{
		while (*in)
		{
			*out = *in;
			out++;
			in++;
		}
	}

The first looks like it should be more efficient and would be on some machines
with the right compiler, However in LightSpeed C, the second is drastically
more efficient due to the order the operations are carried out in. It gets
even better if the pointers are optimised into registers or explicitly copied
into registers.

Randy Buckland
rcb@ncsuvx.ncsu.edu