feat: hash_fr in assembly + removed Utils #794
Conversation
|
|
||
| } | ||
| {{ end }} |
There was a problem hiding this comment.
We do not need to close this block and open again a few lines from here. In both cases it checks len(CommitmentConstraintIndexes) > 0
| let b0 := mload(mPtr) | ||
|
|
||
| // [b0 || one || dst || sizeDomain] | ||
| // <-64bytes -> <- 1 byte each -> |
| error_sha2_256() | ||
| } | ||
|
|
||
| let b0 := mload(mPtr) |
There was a problem hiding this comment.
Cannot we instead store the output of the next SHA2 call to mPtr+0x20. Then b0 will be at mPtr and b1 at mPtr+0x20. Dunno if would help with gas.
There was a problem hiding this comment.
normally b0 is not needed from here, it is only used to populate the first entry of the next hash. So I leave it in mPtr and use it directly for the next hash
| // the result is then 2**(8*16)*mPtr[32:] + mPtr[32:48] | ||
| res := mulmod(mload(mPtr), bb, r_mod) // <- res = 2**128 * mPtr[:32] | ||
| offset := add(mPtr, 0x10) | ||
| for {let i:=0} lt(i, 0x10) {i:=add(i,1)} // mPtr <- [xx, xx, .., | 0, 0, .. 0 || b2 ] |
There was a problem hiding this comment.
Not sure about optimisations. Would unrolling the loop help? Otherwise there are implicit checks for overflow when doing add, no?
There was a problem hiding this comment.
I could try to unroll it. But I might get a 'stack too deep' error if the function is too big
@julien-marchand there might be some conflicts with the 'calldataisation' :/