List of optimizations for Stage 1

[ray6080]

This issue keeps track of possible performance optimization opportunities we have right now.

• nulls in the query processor. add slow/fast path design for nulls, Update the expression evaluation to minimize branching given NULLs #95， Avoid unflat null checks and do memcpy for Flat and Unflat vector operands in expressions #152
• remove the data type switch/cases in vector operation execute function by templation.
• special casing for skipping of selectedPos accesses in the query processor when the selectedPos is 0, 1, 2, ... k.
• null zonemap per page in storage to indicate if there is a null in the page or not.
• copy vs. frame pin for reading from storage.
• add the zonemap index and predicate pushdown to filter out useless tuples from storage.
• Parallelize finalization of the hash table in hash join #177
• Evaluate the overhead caused by virtual function calls for readNodeAndLabel in ValueVector #172
• optimize Metadata design for in-memory #149
• Optimize layout for storing unstructured property #134
• Look into changing locking in SCAN_NODE_ID to use CAS and busy spinning #132
• Consider inplace vector operations and result vector reuse in Expressions #121
• Unnecessary Scans/Reads -> Extra I/O #112
• Microbenchmark on computing hashes in the hash table build phase #67
• OrderBy optimizations #632

[ray6080]

This is outdated now.