Avoid unnecessary copying and dynamic memory allocations #4261
Conversation
Co-authored-by: Chenna Keshava B S <ckbs.keshava56@gmail.com>
| if (sig.size() == 0) | ||
| return false; |
There was a problem hiding this comment.
I don't think this added check is needed as this is taken care of by ecdsaCanonicality below.
There was a problem hiding this comment.
Agreed, I'll remove it.
| PublicKey publicKey_; | ||
| uint256 suppression_; | ||
| Proposal proposal_; | ||
| boost::container::static_vector<std::uint8_t, 72> signature_; |
There was a problem hiding this comment.
I have a hypothetical question. If the size of the signature_ variable was a multiple of 32 bytes, then
can we use a ripple::base_uint<33> for storing the signature_ data? Will that be slower than boost::static_vector?
Is one of them faster than the other?
There was a problem hiding this comment.
ripple::base_uint<33> would not work as it specifies 33 bits (you need 72 bytes or 72 x 8 bits), and it is designed to store uint32_t not an array of bytes. Also base_uint doesn't manage a size, so you'd have to keep track separately of how long the signature is. static_vector is the right choice here!
There was a problem hiding this comment.
I see okay, thanks for the clarification.
This commit implements an optimization that @chennakeshava1998 and I identified; this commit reduces unnecessary dynamic memory allocations when processing proposals.