Description
LLVM and MLIR programs can gain high-performance automatic differentiation for scientific computing, optimization, and machine-learning workloads.
This is compiler and research tooling for developers. Generated derivatives and transformed code should be validated against known results before being trusted in scientific or production calculations.