Fused stages

Stage	Description
LorenzoQuantStage	Fused float predictor + quantizer (1-D/2-D/3-D)
AdaptiveLorenzoStage	Per-tile adaptive multi-order Lorenzo + centering (FSZ prediction stage)
GInterpStage	Multi-level spline interpolation predictor + quantizer (3-D, cuSZ-Hi port)
BitplaneRZEStage	Fused bitplane transpose + zero-group RZE — FZ-GPU's lossless encoder
SZxStage	SZx ultrafast EB compressor — per-block constant/non-constant classification + fixed-length residuals (whole compressor)
SZpStage	SZp / fZ-light — quantize + 1-D Lorenzo delta + fixed-length bitpack (whole compressor)
DifferenceStage (negabinary-fused)	First-order difference with inline negabinary encoding (DifferenceStage<T, TOut> where TOut != T)

‍Note: DifferenceStage with a TOut != T template argument fuses a negabinary encoding step into the differencing kernel (equivalent to DifferenceStage<T> followed by NegabinaryStage, but in a single pass). This stage lives in modules/predictors/diff/, not in modules/fused/, but is listed here so all fused-operation options appear in one place.