Input SQL Query
The pipeline begins with a standard SQL query. GenDB generates instance-optimized C++ code specifically tailored to execute this query on your data and hardware.
Hardware & Data Profiling
The Workload Analyzer agent profiles target hardware capabilities, scans data statistics, and identifies table roles and column access patterns to inform downstream decisions.
Columnar Storage Layout
The Storage Designer agent creates an optimized columnar storage layout with type-specific encodings, sort orders, and indexes tailored to the workload profile.
Physical Query Plan
The Code Generator agent produces a physical execution plan defining the scan strategy, filter pipeline, join order, aggregation operators, and parallelism configuration.
Generated C++ Code
The agent generates complete, compilable C++ code that implements the query plan. Each iteration is a self-contained program using mmap, SIMD intrinsics, and OpenMP parallelism.
Iterative Self-Optimization
The Optimizer agent analyzes execution profiles and applies targeted strategies across multiple iterations — improving data loading, computation, memory access, and parallelism.
Benchmark Results
Final performance comparison of GenDB's generated code against established database systems on this query, measured on identical hardware.