From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-prs-return-24384-listarch-gcc-prs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 26293 invoked by alias); 12 Nov 2002 19:26:21 -0000
Mailing-List: contact gcc-prs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Archive: <http://gcc.gnu.org/ml/gcc-prs/>
List-Post: <mailto:gcc-prs@gcc.gnu.org>
List-Help: <mailto:gcc-prs-help@gcc.gnu.org>
Sender: gcc-prs-owner@gcc.gnu.org
Received: (qmail 26272 invoked by uid 71); 12 Nov 2002 19:26:19 -0000
Resent-Date: 12 Nov 2002 19:26:19 -0000
Resent-Message-ID: <20021112192619.26271.qmail@sources.redhat.com>
Resent-From: gcc-gnats@gcc.gnu.org (GNATS Filer)
Resent-Cc: gcc-prs@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Resent-Reply-To: gcc-gnats@gcc.gnu.org, hartwood@crosslink.net
Received: (qmail 24697 invoked by uid 61); 12 Nov 2002 19:21:41 -0000
Message-Id: <20021112192139.24696.qmail@sources.redhat.com>
Date: Tue, 19 Nov 2002 18:05:00 -0000
From: hartwood@crosslink.net
Reply-To: hartwood@crosslink.net
To: gcc-gnats@gcc.gnu.org
X-Send-Pr-Version: gnatsweb-2.9.3 (1.1.1.1.2.31)
Subject: other/8551: altivec vec_cmplt generates wrong instruction
X-SW-Source: 2002-11/txt/msg00598.txt.bz2
List-Id: <gcc-prs.sourceware.org>


>Number:         8551
>Category:       other
>Synopsis:       altivec vec_cmplt generates wrong instruction
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Tue Nov 12 11:26:17 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Rick Cellucci
>Release:        gcc version 3.3 20021104 (experimental)
>Organization:
>Environment:
Configured with: ../gcc3.3/configure --prefix=/usr/local/darwin --enable-altivec --enable-languages=c --with-cpu=7400 --target=powerpc-apple-darwin6.1
>Description:
Both vec_cmplt and vec_cmpgt generate the same vcmpgtfp instruction. vec_cmplt should generate vcmpgtfp and switch the second and third arguments, which is not happening. Occurs at any optimization level. I assume short and int versions may be broken as well, I only have tested vector floats.
>How-To-Repeat:
void VecCmpltIsBroken (void)
{
	vector float zero    = (vector float) {0.0, 0.0, 0.0, 0.0};
	vector float smaller = (vector float) {1.0, 2.0, 3.0, 4.0};
	vector float bigger  = (vector float) {5.0, 6.0, 7.0, 8.0};
	vector bool int cmp;
			
	cmp = vec_cmplt (bigger, smaller);
	bigger = vec_sel (bigger, zero, cmp);
	PrintFloatVector (bigger); printf ("\n");

	cmp = vec_cmpgt (bigger, smaller);
	bigger = vec_sel (bigger, zero, cmp);
	PrintFloatVector (bigger); printf ("\n");
}


    lvx v31,0,r9
    vcmpgtfp v0,v31,v29
    vxor v30,v30,v30
    vsel v31,v31,v30,v0
    vor v2,v31,v31 
    bl L_PrintFloatVector$stub
    li r3,10       
    bl L_putchar$stub
    vcmpgtfp v0,v31,v29
>Fix:
Workaround: Only use vec_cmpgt and switch arguments order in C.
>Release-Note:
>Audit-Trail:
>Unformatted: