Hello, On Feb 7, 2011, at 7:43 PM, Jean-Michel Pouré - GOOZE wrote: > Le lundi 07 février 2011 à 17:25 +0100, Peter Stuge a écrit : >> Maybe more extensive testing could be done. Higher precision timing, >> and say 1000 or 10000 cycles. > > Okay, I understand what you mean. Timing is different and depends. But > in no way the generation reaches 2m38. It sure did, "time pkcs15-init -G" said so.
> If generation time depends, I would prefer that all timing may be > removed from the information on cards. The procedure was definitely not trying to be scientific. I got curious after seeing that RutokenECP [1] took "forever" (8..10 minutes) to generate a 2048b key because it seems to lack a RSA co-processor (GOST is fast though), so I took two other random cards I knew I could easily re-initialize and did two or three tests, observing that the results were "in the same timespan class" and wrote down one of the results, that was probably close to "the average". The other two cards happened to be MyEID [2] and Feitian [3]. These results are empirical observations of executing set of scripts on OS X 10.6.6 with OpenSC trunk. Such scripts should be consolidated to somewhere in OpenSC repository for both performance as well as regression testing [4]. I agree that any serious "performance testing" must be done on a well defined machine and to specify the environment a bit more (used reader and software versions etc). But the numbers as they were are OK for a "what to expect" number. [1] http://www.opensc-project.org/opensc/wiki/AktivRutokenECP#Speed [2] http://www.opensc-project.org/opensc/wiki/MyEID#Speed [3] http://www.opensc-project.org/opensc/wiki/FTCOSPK01C#Speed [4] http://www.opensc-project.org/pipermail/opensc-devel/2011-January/015797.html -- @MartinPaljak.net +3725156495 _______________________________________________ opensc-devel mailing list opensc-devel@lists.opensc-project.org http://www.opensc-project.org/mailman/listinfo/opensc-devel
