numr 0.5.2

High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/WebGPU)
Documentation
1
2
3
4
5
//! Trait definitions for sparse QR factorization
//!
//! Sparse QR uses free functions per backend (sparse_qr_cpu, sparse_qr_cuda, etc.)
//! rather than a trait-based dispatch pattern, because the CPU implementation
//! operates on extracted f64 data while GPU backends will need native kernels.