On Fri, Dec 12, 2008 at 10:22, Ancoron Luciferis <[email protected]> wrote: > Is the vpnxxxx capable of accelerating the original Rijndael with a > block size of 256 bit (AES is only specified with 128 bit block size)? Part of the speedup in ASICs like the HiFN 7955 (the chip the vpn14x1 is built from) is due to a reduction of flexibility. It is doubtful that they implemented the extended Rijndael functionality; even if they did, you'd have to make sure the driver supported flexible block sizes as well. If it's actually 256-bit _key_ sizes you're looking for, you're in luck, but your 256/256 comment below seems to indicate you know the difference.
> I see that we are using this very often in our code and accelerating > that would give us the most speed-up. SHA-256 would also be nice but > that isn't a major issue regarding to CPU time. The Rijndael 256/256 is > the one that causes most CPU time at some of out applications here. This is why when engineering a system we avoid stepping outside the standards boundaries. Unless you're covering more than 256EB of data with a single key or use a poorly-conceived mode (i.e. ECB), the birthday paradox (which is what the increased block size generally guards against) won't come into play. Regardless, the vpn14x1 is the cheapest add-on accelerator I've found on the market, by an order of magnitude and more. The best way to find out if it works with your edge cases is to get one and try it out; at the current pricing even a private user can afford to do so. RB _______________________________________________ Soekris-tech mailing list [email protected] http://lists.soekris.com/mailman/listinfo/soekris-tech
