Don't waste resources updating hashFinalSproutRoot in src/main.cpp when loading block index #248

Hush nor any Hush Smart Chain has ever had any sprout history in their blockchains, so there is no sense in spending any resources updating hashFinalSproutRoot when loading the block index. From src/main.cpp LoadBlockIndexDB function :

    // Fill in-memory data
    BOOST_FOREACH(const PAIRTYPE(uint256, CBlockIndex*)& item, mapBlockIndex)
    {
        CBlockIndex* pindex = item.second;
        // - This relationship will always be true even if pprev has multiple
        //   children, because hashSproutAnchor is technically a property of pprev,
        //   not its children.
        // - This will miss chain tips; we handle the best tip below, and other
        //   tips will be handled by ConnectTip during a re-org.
        if (pindex->pprev) {
            pindex->pprev->hashFinalSproutRoot = pindex->hashSproutAnchor;
        }
    }

This iterates over every single block in the history of the chain and then updates a value that is never used. It's the only operation in this loop. This makes full node startup times slower and slower as more blocks are in the chain. We can likely comment out the above code with no ill effect and make full node startup times faster.

To test this change :

Verify a full sync can be done
Verify a node can be stopped, restarted and do a partial sync
Create an HSC with default equihash PoW and verify a block can be mined
Verify you can do a full sync with the HSC
Verify you can do a partial sync with the HSC

For extra credit, you can look at the difference in how long it takes the message finished loading blocks to be printed to STDERR on master vs this code change, to see what kind of performance boost we get.

To test this change :

Verify a full sync can be done

Verify a node can be stopped, restarted and do a partial sync

Create an HSC with default equihash PoW and verify a block can be mined

Verify you can do a full sync with the HSC

Verify you can do a partial sync with the HSC

All checks passed.

For extra credit, you can look at the difference in how long it takes the message finished loading blocks to be printed to STDERR on master vs this code change, to see what kind of performance boost we get.

It is nearly instant to begin with, I got a finished loading blocks message within a second on master and I am not sure what performance boost I can test here, could you clarify?

@onryo are you talking about loading blocks on HUSH mainnet or an HSC? For an HSC wih just a few blocks, it's going to be instant. For HUSH mainnet, which currently has over 1.1 million blocks, it often takes at least 1 or 2 minutes, depending on CPU speed, in my experience.

@onryo if things work for you without that code, please send a PR to the dev branch with that code deleted. I am not sure what performance change it will give, but at least we are getting rid of sprout code we don't need.

@onryo are you talking about loading blocks on HUSH mainnet or an HSC? For an HSC wih just a few blocks, it's going to be instant. For HUSH mainnet, which currently has over 1.1 million blocks, it often takes at least 1 or 2 minutes, depending on CPU speed, in my experience.

To be honest I was talking about HUSH, not HSC. Takes me 1-2 seconds. 😶

done, closing