Path: utzoo!utgpu!utstat!jarvis.csri.toronto.edu!mailrus!cornell!uw-beaver!rice!sun-spots-request
From: flee@shire.cs.psu.edu (Felix Lee)
Newsgroups: comp.sys.sun
Subject: Strength reduction flub in Sun-4 C compiler
Message-ID: <8902110038.AA10231@shire.cs.psu.edu>
Date: 22 Feb 89 00:52:43 GMT
Sender: usenet@rice.edu
Organization: Sun-Spots
Lines: 37
Approved: Sun-Spots@rice.edu
Original-Date: Fri, 10 Feb 89 19:38:10 EST
X-Sun-Spots-Digest: Volume 7, Issue 162, message 2 of 11

Machine Type:   Sun 4/260S
O/S Version:    SunOS 4.0
Organization:   Computer Science Department, The Pennsylvania State University
                333 Whitmore Laboratory, University Park, PA   16802

Description:
	Compile the following code with -O2 optimization:

	main()
	{
		int k, total, x, y;

		total = 0;
		for (k = 3; k > 0; --k) {
			x = total;
			total += 7;
			x = total - x;
			y = x << 2;
			printf("%d\n", y);
		}
	}

	Instead of printing 28, 28, 28 as you would expect, it produces
	0, 28, 56. 

	Looking at the assembly code, the compiler is doing strength
	reduction in the outer loop, incorrectly:
		%i4 = 0;
		for (k = 3; k > 0; --k) {
			printf("%d\n", %i4);
			%i4 += 28;
		}
	%i4 is a register that tracks the value (total*4).

	Despite being wrong, the strength reduction is of dubious value...
--
Felix Lee