On 10/13/2013 12:14 AM, Andy Polyakov wrote:
>> In OpenWrt we have a build error in the openssl sha256 code on MIPS I.
> For reference, I've never tested MIPS modules on pre-III, and therefore
> feedback for earlier generations is appreciated. It was tested to
> compile for MIPS II though, for IRIX account.
>> /usr/bin/perl asm/sha512-mips.pl o32 sha256-mips.S
>> mipsel-openwrt-linux-uclibc-gcc -O2 -pipe -mtune=r5000 -fno-caller-saves
>> -fhonour-copts -Wno-error=unused-but-set-variable -msoft-float -fpic -c
>>   -c -o sha256-mips.o sha256-mips.S
>> sha256-mips.S: Assembler messages:
>> sha256-mips.S:1960: Error: opcode not supported on this processor: mips1
>> (mips1) `bnel $5,$23,.Loop'
>> make[5]: *** [sha256-mips.o] Error 1
> Well, if you -mtune=r5000 you probably are not actually targeting MIPS
> I, and could therefore specify even -mips3 on compiler command line.
> But, yes, as you point out it's appears to be declared obsolete in the
> latest specification, so that switching to bne is appropriate. But just
> replacing it with bne is not necessarily appropriate because instruction
> is delay slot to b*l is executed conditionally. Also note that there are
> more b*l instructions, in crypto/bn/asm. I'll see what is appropriate to
> do, meanwhile specify at least -mips2 on command line.
> On side note, last update to MIPS modules (in 1.0.2 and forward) added
> support for latest MIPS[32|64]R2 and SmartMIPS extensions. If you happen
> to have such CPUs, it would pay off to specify matching command-line
> option(s).

I do not have this hardware, I just saw that some automatic builds are
failing and wanted to inform the author of that code, the build was done
by cross compiling. The Mips devices I have here are all MIPS32 R1 and
R2. Doing some buildtests with these compiler optimizations that are
causing problems here is no problem for me.

OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev@openssl.org
Automated List Manager                           majord...@openssl.org

Reply via email to