From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43)
	id 1KiFl8-0005fy-AT
	for qemu-devel@nongnu.org; Tue, 23 Sep 2008 17:49:26 -0400
Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43)
	id 1KiFl7-0005fT-7i
	for qemu-devel@nongnu.org; Tue, 23 Sep 2008 17:49:25 -0400
Received: from [199.232.76.173] (port=59826 helo=monty-python.gnu.org)
	by lists.gnu.org with esmtp (Exim 4.43) id 1KiFl7-0005fO-0w
	for qemu-devel@nongnu.org; Tue, 23 Sep 2008 17:49:25 -0400
Received: from csl.cornell.edu ([128.84.224.10]:1352 helo=vlsi.csl.cornell.edu)
	by monty-python.gnu.org with esmtp (Exim 4.60)
	(envelope-from <vince@stanley.csl.cornell.edu>) id 1KiFl7-0004zJ-7P
	for qemu-devel@nongnu.org; Tue, 23 Sep 2008 17:49:25 -0400
Received: from stanley.csl.cornell.edu (stanley.csl.cornell.edu
	[128.84.224.15])
	by vlsi.csl.cornell.edu (8.13.4/8.13.4) with ESMTP id m8NLnDYb075786
	for <qemu-devel@nongnu.org>; Tue, 23 Sep 2008 17:49:18 -0400 (EDT)
Date: Tue, 23 Sep 2008 17:49:13 -0400 (EDT)
From: Vince Weaver <vince@csl.cornell.edu>
Subject: re: [Qemu-devel] [5283] Use the new concat_tl_i64 op for std and stda
Message-ID: <20080923174506.X10311@stanley.csl.cornell.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed
Reply-To: qemu-devel@nongnu.org
List-Id: qemu-devel.nongnu.org
List-Unsubscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.gnu.org/pipermail/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: qemu-devel@nongnu.org

Hello

the aforementioned patch broke the spec2k crafty benchmark on 
sparc32plus-linux-user (I'm running on x86_64 but I don't think it makes a 
difference).

As always, the problem is when there are high bits (above 32) set when 
concating.

The old code generated:
    mov_i32 tmp11,tmp0
    mov_i32 tmp1,loc4
    movi_i64 tmp12,$0xffffffff
    and_i64 tmp10,tmp1,tmp12
    movi_i64 tmp12,$0x20
    shl_i64 tmp10,tmp10,tmp12
    movi_i64 tmp12,$0xffffffff
    and_i64 tmp2,tmp11,tmp12
    or_i64 tmp2,tmp2,tmp10
    qemu_st64 tmp2,loc5,$0x0

The new code generates:
    movi_i64 tmp12,$0x20
    shl_i64 tmp10,loc4,tmp12
    or_i64 tmp2,tmp0,tmp10
    qemu_st64 tmp2,loc5,$0x0

which generates wrong values because the low value being concatenated 
isn't masked/truncated to 32-bits first.

Vince