Skip to main content

xlog_induce/
score.rs

1//! Batched topology scoring orchestration over request-local indexes.
2//!
3//! The native bounded exact-induction path drives the `kernels/ilp_exact.cu`
4//! kernels that score all `(left, right)` pairs for all four topologies in a
5//! single batched GPU pass, avoiding per-candidate host transfers.