RandomX commitments with double-hashing

[tevador]

Related to this proposal: monero-project/monero#8827

The commitment is calculated from the hash and its input (i.e. the block header). The idea is that the output of randomx_calculate_hash will be appended to the block header and the difficulty will be calculated based on the output of randomx_calculate_commitment, which is very fast.

[SChernykh]

Looks fine to me. What's the status of this PR? Is it the final version?

[tevador]

I'm not planning any modifications to this PR unless something comes up during the review. The changes are fairly small, so it's mostly about the concept.

One thing to note is that batched hashing using the randomx_calculate_hash_next function will become slightly more complicated due to the need to recall the previous hash input when the v1 result is ready.

[SChernykh]

One thing to note is that batched hashing using the randomx_calculate_hash_next function will become slightly more complicated due to the need to recall the previous hash input when the v1 result is ready.

That wouldn't be an issue for mining software because it only needs to remember the previous nonce value (4 bytes). But yes, randomx_calculate_hash_next will need two more parameters for the previous input.

[bitcartel]

Hi, this can help prevent user error when calling the library.
assert(v1_in != v2_out);

[tevador]

It will work perfectly fine with v1_in == v2_out.

[bitcartel]

Thanks, good to know.

[tevador]

I renamed the new function to randomx_calculate_commitment because it provides a commitment to a RandomX hash value. The "v2" naming was not a good choice because this change is orthogonal to future VM changes that might need to be versioned.