On Fri, Sep 24, 2004 at 12:38:51AM -0400, Theodore Ts'o wrote:
> I'm also a bit concerned about how much time AES takes over the
> cut-down MD4, as this may affect networking benchmarks.  (And we don't
> need super-strength crypto here.)

Oh,

openssl speed md4 aes shows:

type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
md4              10708.72k    38240.96k   111170.47k   215872.85k 296828.93k
aes-128 cbc      32121.81k    32678.31k    33119.49k    33221.29k 33210.59k
aes-192 cbc      27915.92k    27868.52k    28418.08k    28677.12k 28721.15k
aes-256 cbc      24599.57k    25142.38k    25381.80k    25474.88k 25392.46k

Since we're using small blocks.

Attached is the patch with the problems Ted pointed out.

JLC