[cryptography] Algorithm that can be *least* optimized on CPUs?

Paul Hoffman paul.hoffman at vpnc.org
Mon Sep 1 10:49:08 EDT 2014


On Sep 1, 2014, at 1:15 AM, Rob Stradling <rob.stradling at comodo.com> wrote:

> On 31/08/14 19:53, Paul Hoffman wrote:
>> Greetings. I want to use "openssl speed" as a very loose proxy for "how fast is this CPU right now". (Note the use of the word "very" there.) I also want to test systems that are running in VMs. However, if I'm comparing across CPU families, OpenSSL might have some optimizations for some algorithms (like AES) that work better or worse depending on the features of the CPU.
>> 
>> If I were to pick one algorithm that is least likely to be optimized past normal C optimization, which would it be?
> 
> Paul, if you build OpenSSL with the "no-asm" configuration option, presumably none of the algorithms will be optimized past normal C optimization.  Does this help?

Yep, thanks. Also, I got an off-line note that no one had ever figured out how to optimize TripleDES at all, so that was a safe one to use as a reference.

--Paul Hoffman


More information about the cryptography mailing list