From mboxrd@z Thu Jan 1 00:00:00 1970 From: Baruch Siach Date: Tue, 28 Jun 2011 07:23:24 +0300 Subject: [Buildroot] how to tune powerpc e500v2 cpus in BR In-Reply-To: <4E08E6DB.4070606@rossvideo.com> References: <4E04C9EE.4050206@rossvideo.com> <20110627035939.GB3046@jasper.tkos.co.il> <4E08E6DB.4070606@rossvideo.com> Message-ID: <20110628042324.GA3154@jasper.tkos.co.il> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Matias, raymod, On Mon, Jun 27, 2011 at 04:23:55PM -0400, Matias Garcia wrote: > On 11-06-27 04:10 PM, raymond zhao wrote: > >On Sun, Jun 26, 2011 at 11:59 PM, Baruch Siach wrote: > >>Hi Raymond, Matias, > >>I'm also using the CodeSourcery toolchain for the e500v2 target (P1011), and I > >>figured out that I must pass -te500v2 to gcc at link time as well. Since there > >>is no generic way to add LDFLAGS options in Buildroot, I've posted the patch > >>at http://lists.busybox.net/pipermail/buildroot/2011-June/043668.html. Did you > >>find a better solution? > >> > >I took a look at the CodeSourcery. Looks it doesn't support uCLibc. Yes. I'm also trying to build a uClibc toolchain using crosstool-ng. No luck yet. You may try the Aboriginal Linux toolchain (http://landley.net/aboriginal/downloads/binaries/), although it's based on an old gcc version (4.2.1). > changbing: That's correct. You may find an older toolchain from them > that uses uClibc, but you'll have to specify it yourself in the > Buildroot config. I found a very small difference in size (<2MB) > between uclibc and eglibc filesystems, but I'm using 32MB of flash > just for the OS, so small is relative. The greatest advantage of uClibc is that it behaves better when doing static linking. I use it to generate Busybox based initramfs, and for this use case the 2MB size reduction is a big difference. > Baruch: I use the external toolchain wrapper introduced just before > 2011.05, but I had to patch it to add arbitrary options. I posted a > couple of patches here, one that introduces a new option for > selecting the sysroot when compiling the external wrapper, and > another that simply passes the BR2_TARGET_OPTIMIZATIONS to the > external wrapper, but they didn't get much attention. Can this solve the link time -te500v2 issue? I'll look into this. baruch > I prefer the first method (new config option) because you can choose > not to pass the optimizations at some later point for whatever > reason, and thus use the wrapper only to select your sysroot. > > Cheers, > Matias -- ~. .~ Tk Open Systems =}------------------------------------------------ooO--U--Ooo------------{= - baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -