tpcds-benchmark

You can benchmark TPC-DS queries by automating data generation, running 99 queries across DuckDB CPU, Legacy Sirius, and Super Sirius, and compiling comparative results to evaluate engine performance.

Yes, you can compare GPU vs CPU performance by controlling GPU behavior through environment variables and config files during the TPC-DS benchmark execution across Legacy Sirius, Super Sirius, and DuckDB.

TPC-DS benchmark validation identifies failures, timeouts, and GPU fallbacks by parsing log files and timing outputs after executing the 99 queries across the configured SQL engines.

The best way to run TPC-DS data generation at multiple scales is using the provided scripts to orchestrate dataset creation before executing the 99 benchmark queries across your target engines.

Yes, you need specific environment variables and config files to run TPC-DS benchmarks on Super Sirius, as these settings control GPU versus CPU behavior and ensure proper execution across engines.

Inconsistent TPC-DS benchmark results between DuckDB and Sirius engines often stem from GPU fallbacks or timeouts, which the validation process detects by parsing logs and timing outputs.