g11tech
g11tech
> OK had to take some time to find some gotcha's here, but this is ready for a first round of reviews. I'm not entirely sure about the implemented API...
@jochem-brouwer i have digged into the PR, looks incredible :fire: , will try next running it to get a feel of the `walkTrie` helper :+1:
I think i will need to fix the tests since the pubsub reference has changed, let me update and push
@dapplion ready for merge!
will investigate :+1:
@dapplion just to confim, the `endianness` in ssz serialization is `little`?
1. as first step, it seems that dataview based to_bytes is at least 2 times faster than the current to_bytes loop for >32 bit value for unit64 and 3.5 times...
@dapplion got 88% speedup on normal uint tree backed (will need to check on > 2^32 number increment test but should be 30% plus over there), marginal 4% better performance...
reworking here: https://github.com/ChainSafe/lodestar/pull/4604
will be picking this up