Re: toolchain/45673: gcc on i386 current generates bad code for assembler code in lock.h

["Kai-Uwe Eckhardt" <kuehro%gmx.de@localhost>]

The following reply was made to PR toolchain/45673; it has been noted by GNATS.

From: "Kai-Uwe Eckhardt" <kuehro%gmx.de@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: toolchain/45673: gcc on i386 current generates bad code for 
 assembler code in lock.h
Date: Sun, 11 Mar 2012 10:06:54 +0100

 Hi,
 
 it looks like a problem in the toolchain. The gcc in 6.0_BETA and current
 i386 is built with --with-march=i486 --with-tune=nocona. Nocona has the
 64 bit registers available and this leads to wrong code. It also
 fails on amd64 using -m32 where -mtune=nocana is default as well. 
 Running the test case with -m32 -march=i486 -mtune=pentiumpro works fine.
 I propose fixing x86/lock.h on i386 and amd64 by using q for the register
 usage directive. Using nocona as default tuning in i386 should be 
 reconsidered. I suggest pentiumpro which is a good default value for
 all processors in my experience.   
 
 Kai-Uwe
 -- 
 --