From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from penguin.netx4.com (embeddededge.com [209.113.146.155]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTP id D0C892BF0B for ; Wed, 10 Nov 2004 04:14:18 +1100 (EST) In-Reply-To: <20041109163726.30776.qmail@web53808.mail.yahoo.com> References: <20041109163726.30776.qmail@web53808.mail.yahoo.com> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <22B5F92B-3273-11D9-8E74-003065F9B7DC@embeddededge.com> From: Dan Malek Date: Tue, 9 Nov 2004 12:16:46 -0500 To: annamaya Cc: linuxppc-embedded@ozlabs.org Subject: Re: MPC8xx binaries running on a MPC82xx board List-Id: Linux on Embedded PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Nov 9, 2004, at 11:37 AM, annamaya wrote: > I was under the assumption that the MPC8xx core was > different from the MPC82xx core (the 603e?). Yes, it is different, but the instructions are all the same. > ...... But I was > surprised to find that the MPC8xx binaries were > running fine on a MPC82xx binaries, even if they were > dynamically linked and were linking to the MPC82xx > libraries. What's going on here? Partly luck, and partly that the instructions are the same. The luck part is you aren't using floating point. The 8xx doesn't have hardware floating point, but the 82xx generated code may have some instructions. It also depends upon the tools you are using. If you were using different glibc versions, your luck would have run out. From a user space application perspective, there are exactly two differences between 8xx and 82xx. One is floating point and the other is cache line size. The 8xx software will run fine on the 82xx, but 82xx may not work properly on 8xx. > ...... Can I just use the > 8xx toolkit to work with the 82xx? An 8xx toolkit will always work on 82xx, but depending upon the way it was built, you will not get the maximum performance of the 82xx. Some tool vendors generate toolkits (and libraries) specifically for the 8xx, where the floating point is all emulated in software. If you were looking for floating point performance, you would be very disappointed with this on the 82xx. > ....... There must be SOME > differences between these two architectures. Can > someone here explain this to me? There are lots of differences in the kernel programming model, but only the two I mentioned for user space applications. In fact, across the entire PowerPC architecture you will only see floating point, Altivec, SPE, and some cache line operation/implementation differences at the user level. Depending upon your software requirements and implementation, you are likely able to run many of the same binaries across all of the PowerPC cores. I guess we are rewriting the archives now? :-) -- Dan