The patched worked great and clang performance is now pretty much on par 
with gcc. It was also necessary to add "-msse4" to the compile flags after 
applying your patch. The relevant compile scenarios are now

$ *g++-5.1 *-Wa,-q -DNDEBUG -O2 -arch x86_64 -DCRYPTOPP_DISABLE_ASM -pipe 
-c ...
$ *clang++* -DNDEBUG -O2 -maes -mpclmul -msse4 -DCRYPTOPP_DISABLE_ASM -pipe 
-c ...
$* icpc* -DNDEBUG -O2 -DCRYPTOPP_DISABLE_ASM -pipe -c ...

Intel:    ~1.7 GiB/second
Clang: ~3.5  GiB/second
GNU:   ~4.1 GiB/second

I still find it surprising that the Intel compiler is being left in the 
dust here. Maybe Intel is trying to tell me I need to directly use the 
Intel Performance Primitives Crypto Library for fast AES with their 
compiler.

On Friday, April 24, 2015 at 1:07:07 PM UTC-7, Gabriel Hackebeil wrote:
>
> Mouse,
>
> I just came across your post from a few months ago about this exact 
> compilation issue (
> https://groups.google.com/forum/#!searchin/cryptopp-users/Cannot$20compile$20crypto$2B$2B$20%28camellia%29/cryptopp-users/TfkVPnljrzg/DwbHIcbD8O4J).
>  
> Trying out your patch now.
>

-- 
-- 
You received this message because you are subscribed to the "Crypto++ Users" 
Google Group.
To unsubscribe, send an email to [email protected].
More information about Crypto++ and this group is available at 
http://www.cryptopp.com.
--- 
You received this message because you are subscribed to the Google Groups 
"Crypto++ Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to