Second Attempt of VAES #186
Conversation
|
Patch will change hash behavior w/ or w/o VAES. Bascially, it increases the batch size to 2x128bits for large enough input and add another two rounds of Without VAES: With VAES: |
| } | ||
| } | ||
|
|
||
| pub(crate) fn shuffle_and_add_vec256(base: Vector256, to_add: Vector256) -> Vector256 { |
There was a problem hiding this comment.
Here and below, these should probably have an inline annotation
| @@ -160,6 +210,9 @@ impl Hasher for AHasher { | |||
| self.hash_in(value.convert()); | |||
| } else { | |||
| if data.len() > 32 { | |||
| if data.len() > 128 { | |||
There was a problem hiding this comment.
This is sub-optimal. It's handling the >128 case but then the >64 case below is only for 65-128, which could be specialized be more efficient.
I would be tempted to avoid the additional if check if possible as branch mispredictions are more common and costly than show up in benchmarks where the data is predictable. But I don't have a good sense on how much is lost operating on 4 blocks at a time as opposed to 8.
No description provided.