From mboxrd@z Thu Jan 1 00:00:00 1970 From: Albert ARIBAUD Date: Wed, 24 Aug 2011 11:41:42 +0200 Subject: [U-Boot] [PATCH v5] ARM: Add Calxeda Highbank platform In-Reply-To: <4E54B862.2070000@aribaud.net> References: <1313932575-31456-1-git-send-email-robherring2@gmail.com> <4E5118C2.7010404@aribaud.net> <4E515C5D.3030902@gmail.com> <4E53B9F9.7020109@gmail.com> <4E54B862.2070000@aribaud.net> Message-ID: <4E54C756.1020609@aribaud.net> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Le 24/08/2011 10:37, Albert ARIBAUD a ?crit : > Le 23/08/2011 16:32, Rob Herring a ?crit : >> Albert, >> >> On 08/21/2011 02:28 PM, Rob Herring wrote: >>> On 08/21/2011 09:40 AM, Albert ARIBAUD wrote: >>>> Hi Rob, >>>> >>>> Le 21/08/2011 15:16, Rob Herring a ?crit : >>>>> From: Rob Herring >>>>> >>>>> Add basic support for Calxeda Highbank platform. Only minimal support >>>>> with >>>>> serial and SATA are included. >>>>> >>>>> Signed-off-by: Jason Hobbs >>>>> Signed-off-by: Rob Herring >>>>> Cc: Albert ARIBAUD >>>>> --- >>>> >>>> This applies to u-boot-arm/master, with three empty lines at end of new >>>> files removed, but the compilation with ELDK 4.2 fails with: >>>> >>>> uboot at lilith:~/src/u-boot-arm$ ./MAKEALL highbank >>>> Configuring for highbank board... >>>> include/common.h:0: error: bad value (armv7-a) for -march= switch >>>> include/common.h:0: error: bad value (armv7-a) for -march= switch >>>> lib/asm-offsets.c:1: error: bad value (armv7-a) for -march= switch >>>> make: *** [lib/asm-offsets.s] Erreur 1 >>>> arm-linux-size: './u-boot': No such file >>>> >>>> --------------------- SUMMARY ---------------------------- >>>> Boards compiled: 1 >>>> Boards with warnings or errors: 1 ( highbank ) >>>> ---------------------------------------------------------- >>>> >>>> Seems like ELDK 4.2 does not know armv7-a. Is it a requirement that this >>>> board build with this march flag? >>>> >>> >>> With device tree, there are lots of endian conversions. With v7, those >>> conversions are a single inline "rev" instruction vs. 10-20 assembly >>> instructions. Since there is no state maintained in the fdt code, the >>> conversions are done over and over. >>> >>> So there's not a hard requirement, but it is something I will always set >>> and I imagine others with v7 platforms will want to also. I think Tegra >>> prevents doing this change globally. >>> >> >> What do you want to do here? >> >> arch/arm/cpu/armv7/highbank/config.mk can simply be deleted if you don't >> want to compile with armv7-a. > > Well, it is more a question of which toolchain(s) we want supported in > U-Boot. Denx ELDK is one of them I guess :) but now that ELDK 5.0 is > out, I don't know if ELDK 4.2 support is still required. > > (I am assuming that ELDK5.0 has armv7-a support here. Please correct me > if not) > > If ELDK4.2 must be supported, then I would suggest creating a > configuration option to allow the use of armv7-a ISA, then the > arch/arm/cpu/armv7/highbank/config.mk file would select armv7-a ISA > based on this option, and then any ARMv7 compatible board that needs or > wants to use this ISA can set this config option in its config header file. > > If ELDK4.2 support is dropped (at least partially) then no change would > be needed. ... adding Wolfgang, this time... > Wolfgang:, I think that toolchain support for ARM is more your choice > than mine. > >> Rob > > Amicalement, Amicalement, -- Albert.