ryankert01 opened a new pull request, #1478: URL: https://github.com/apache/mahout/pull/1478
### Related Issues Part of #1477 (1/5). ### Changes - [x] Test - [x] Other: benchmark tooling ### Why Before touching the encode path, the branch needs a specification and a performance gate that exist independently of the Rust and CUDA layers. Today the only cross-backend check is the PyTorch reference tests, and phase has no reference at all. ### How - `qumat_qdp.torch_ref.phase_encode`: the pure-PyTorch reference now covers every encoding the engine supports. - `testing/qdp/test_parity.py`: one parametrised test over encoding × engine precision × shape × input location (NumPy, torch CPU, CUDA f64, CUDA f32), each cell checked against `torch_ref`. Known gaps (encodings without an f32 device kernel) are skipped explicitly. - `qdp/qdp-python/benchmark/baseline.py`: capture throughput and latency for every encoding and dtype into JSON, and compare a later build against it with a regression threshold. Numbers are machine-specific and are not committed. No engine code changes. `make parity` (added in 5/5) runs the grid. ## Checklist - [x] Added or updated unit tests for all changes - [x] Added or updated documentation for all changes --- Part of a five-PR series from the same branch, each adding one commit on top of the previous; the diff shown is cumulative against `main`. Series: this. -- 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]
