[openssl-users] OpenSSL 1.0.2h reports speed test results as 0 secs and Infk ops/sec

[Jeff Wieland]

Dennis Clarke wrote:
>
>>>
>> I do build with the no-asm option, and I'm seeing the problem.
>>
>> I'm suspicious that somehow the compiler optimization is generating code
>> that doesn't work quite right on the UltraSPARC 2e.
>
> I have seen this a few times now so I also felt, hrmmm, something not 
> quite right happening on these old slow netra boxes. Which, by the 
> way, make great indestructible DNS servers.
>
> In any case I changed the CFLAGS for the solaris64-sparcv9-cc option in
> Configure thus :
>
>
> "solaris64-sparcv9-cc","cc:
>
> -m64 -xtarget=ultra2e -xarch=sparcvis -xchip=ultra2e -xcache=generic
> -errfmt=error -erroff=%none -errshort=full -xstrconst -xildoff
> -xmemalign=8s -xnolibmil -Xa -xcode=pic32 -xregs=no%appl -xlibmieee
> -mc -g -xs -ftrap=%none -Qy -xbuiltin=%none -xdebugformat=dwarf
> -xunroll=1 -D_TS_ERRNO -D_POSIX_PTHREAD_SEMANTICS
> -D_LARGEFILE64_SOURCE -D_REENTRANT -xdepend -DB_ENDIAN
>
> ::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:
>
> BN_LLONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL 
> BF_PTR:${sparcv9_asm}:dlfcn:solaris-shared:
> -KPIC:-m64 -G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR):::/64",
>
>
> So the objective was to correct the wrong -xarch flag that has been in 
> there for ages and also to get a full debug version which would be 
> easy to single step through.
>
> So that works fine.
>
> So what I will look for is where the time calc is done, single step 
> through that and find out why we get a 0.00sec time.
>
> Dennis
I get the same results compiling with Oracle's build of gcc in 
/usr/sfw/bin (which uses
the Solaris assembler, etc.) and my own build of gcc 3.4.6 (which uses 
the GNU
assembler, etc.).

Have you tried running Oracle's builds of OpenSSL?  They do the same 
thing on the
UltraSPARC 2e:

$ /usr/bin/openssl version;/usr/bin/openssl speed
OpenSSL 1.0.1t  3 May 2016
Doing md2 for 3s on 16 size blocks: 140755 md2's in 0.00s
Doing md2 for 3s on 64 size blocks: 73864 md2's in 0.00s
Doing md2 for 3s on 256 size blocks: 25778 md2's in 0.00s
Doing md2 for 3s on 1024 size blocks: 6695 md2's in 0.00s
...

$ /usr/sfw/bin/openssl version;/usr/sfw/bin/openssl>
OpenSSL 0.9.7d 17 Mar 2004 (+ security fixes for: CVE-2005-2969 
CVE-2006-2937 CVE-2006-2940 CVE-2006-3738 CVE-2006-4339 CVE-2006-4343 
CVE-2006-7250 CVE-2007-5135 CVE-2007-3108 CVE-2008-5077 CVE-2008-7270 
CVE-2009-0590 CVE-2009-2409 CVE-2009-3555 CVE-2010-4180 CVE-2011-4576 
CVE-2011-4619 CVE-2012-0884 CVE-2012-1165 CVE-2012-2110 CVE-2012-2131 
CVE-2012-2333 CVE-2013-0166 CVE-2013-0169 CVE-2014-0224 CVE-2014-3508 
CVE-2014-3511 CVE-2014-3566 CVE-2014-3567 CVE-2014-3568 CVE-2014-3569 
CVE-2014-3570 CVE-2014-8275 CVE-2015-0204 CVE-2015-0286 CVE-2015-0287 
CVE-2015-0288 CVE-2015-0289 CVE-2015-0292 CVE-2015-0293 CVE-2015-1789 
CVE-2015-1790 CVE-2015-3195 CVE-2015-3197 CVE-2015-4000 CVE-2016-0703 
CVE-2016-0704 CVE-2016-0797 CVE-2016-0799 CVE-2016-0800 CVE-2016-2105 
CVE-2016-2106 CVE-2016-2107 CVE-2016-2108 CVE-2016-2176)
Doing md2 for 3s on 16 size blocks: 79067 md2's in 0.00s
Doing md2 for 3s on 64 size blocks: 42773 md2's in 0.00s
Doing md2 for 3s on 256 size blocks: 15316 md2's in 0.00s
Doing md2 for 3s on 1024 size blocks: 4240 md2's in 0.00s
Doing md2 for 3s on 8192 size blocks: 543 md2's in 0.00s
...


They appear to work fine on the other SPARC machines that I can get test 
it on.

--
Jeff Wieland, UNIX/Network Systems Administrator
Purdue University IT Infrastructure Services UNIX Platforms