Hacking Postgres Executor For Performance
summary
Talk documents concrete patches and bigger ideas: expression evaluation is flattened into a single function since v10, ExecProcNode dispatch is replaced by prebound function pointers, Heap AM gets inlinable all-visible page paths, and tuple-deform uses a compact 16-byte attribute layout — these localized changes show 5–25% faster OLAP queries on tpch-like workloads and per-patch gains up to ~5% for exec scan. Future work outlines an opcode-based executor, batch/columnar processing and LLVM/JIT compilation to remove interpreter overhead and enable SIMD/vectorization.
AI-written summary · about
related talks
32mHacking Postgres Executor For Performance
41mPostgreSQL Executor: Executing your execution plan
50mImproving Postgres' Efficiency
34mPostgres 10, Performance, and You
50mPerformance Archaeology
52mWriting fast C code for a modern CPU (and applying it to PostgreSQL)
56mPerformance Improvements in PostgreSQL9 2
46mInvestigating Multithreaded PostgreSQL