On Tue, Sep 6, 2016 at 10:07 AM, Jean-Pierre Münch <[email protected]> wrote: > Am 06.09.2016 um 16:05 schrieb raad via Crypto++ Users: > > Am Dienstag, 6. September 2016 14:08:34 UTC+2 schrieb Jeffrey Walton: >> >> >> Does 'cryptest.exe v' feel like its running any faster? Or more >> importantly, did things feel like they slowed down? We definitely need to >> avoid the latter. > > > I got 3034 ms without the constexpr changes and 3022 ms with them in x64 > release mode. Tried 12 times each on my i7-6700. > > I also had problems with the RDSEED test in 5.6.3 (one of the > *_GenerateBlock functions in rdrand.asm failed most of the time, but not > always), but it works for me now. > > That's because we used to set the retry-threshold for RDSEED fairly low (it > fails in case it doesn't get random bytes in-time and the TRNG isn't the > fastest) and we requested way more bytes than the system was designed for. > We fixed this by increasing the retry threshold in the mean-time.
Yeah, I was not aware until recently from talking with DJ from Intel.... RDRAND is designed to never underflow. If it underflows once, then there's a problem. RDSEED will underflow on occassion, so we have to be prepared for failures. I think we increased the RDSEED retry value 2 or 3 times looking for the "sweet spot". Jeff -- -- 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.
