From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <3AE58FCB.8D4198C3@matrox.com> Date: Tue, 24 Apr 2001 10:38:05 -0400 From: Alexandre Nikolaev Reply-To: Alexandre.Nikolaev@matrox.com MIME-Version: 1.0 To: linuxppc-embedded@lists.linuxppc.org Subject: Re: ppc little-endian port References: <3AE47B3C.8DF8F330@matrox.com> <3AE5595C.74644FE5@raleigh.ibm.com> Content-Type: text/plain; charset=koi8-r Sender: owner-linuxppc-embedded@lists.linuxppc.org List-Id: Sorry, we were not clear enough... Why we did it: We have our own image processing PCI board. Hardware guys decided that since PCI is little endian, it would be better from "their hardware point of view" to run ppc in little-endian as well. We can do nothing about that and just have to run ppc in little endian. Maybe someone else will be on the same boat :) What else we did: We compiled the whole Busy Box using glibc with some modifications of glibc (endian related + strange bug). Also, we compiled the ash of standard source (taken from Montavista) using Dietlibc with no modifications. Then we compiled and run an application under ash shell (yieldt benchmark from Montavista) and saw good results: typically 7 ms and 11 ms thread context switch time regardless of number of threads and thread size, for revisions 2.09 and 2.06 respectfully. Which processors did we test: We ran only 7400, revision 2.06 and 2.09. Alexandre ***************************************** | Alexandre Nikolaev | Software Designer | Matrox Electronic Systems Ltd | Imaging Department | Tel. (514) 8226000 ext. 2323 | E-mail: Alexandre_Nikolaev@matrox.com *****************************************:) ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/