[2.29, 3.18]) in a cached-vs-uncached ablation of XLOG’s own
pipeline, and steady epochs run 2.80× faster than Scallop at matched accuracy.
Quickstart
Install XLOG, write your first program, and run it on the GPU in five minutes.
Language Reference
The full language surface: types, predicates, rules, modules, aggregation, and pragmas.
Probabilistic Programming
Exact inference via GPU knowledge compilation, plus Monte Carlo sampling with confidence intervals.
Architecture
How the compiler, IR stack, and CUDA runtime fit together.