andygrove opened a new pull request, #3213: URL: https://github.com/apache/datafusion-comet/pull/3213
## Summary **This is a work-in-progress PR - not ready for review yet.** This PR optimizes `CometColumnarToRowExec` for complex types (arrays and maps) by: ### Phase 1: Cache offset buffer addresses in Comet vectors - `CometListVector` and `CometMapVector` now cache the offset buffer memory address at construction time - `getArray()`/`getMap()` use `Platform.getInt()` for direct memory access instead of `getOffsetBuffer().getInt()` - This eliminates 2 virtual method calls per offset read (4 total per `getArray()`/`getMap()` call) ### Phase 2: Custom code generation for complex types - `CometColumnarToRowExec` now generates optimized code for `ArrayType` and `MapType` columns - Per-batch caching of offset buffer addresses and child vectors in generated code - Uses `Platform.getInt()` directly in generated code for offset access ### Before (per row): ```java listVector.getOffsetBuffer().getInt(i * OFFSET_WIDTH) // 2 virtual calls + bounds check ``` ### After (per row): ```java Platform.getInt(null, offsetBufferAddress + (long) i * OFFSET_WIDTH) // JIT intrinsified ``` ## TODO - [ ] Phase 3: Reusable wrapper objects (MutableColumnarArray/Map) to eliminate object allocation - [ ] Benchmarking to measure performance improvements - [ ] Additional test coverage - [ ] Handle StructType similarly ## Test plan - [x] `CometArrayExpressionSuite` passes (32 tests) - [x] `CometMapExpressionSuite` passes (2 tests, 2 canceled, 1 ignored) - [ ] Full test suite - [ ] Performance benchmarks 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
