Expand description
Items the other sparse-ir crates build on, outside the semver
guarantees of this crate
These are the machinery of the SVE (its strategies, the discretized
kernel matrices, the truncated SVD), the FPU guard and the thread pool.
Any release may change or remove them; a program that only computes
SVEs does not need them. sparse-ir-basis and sparse-ir re-export
some of them under their own paths (sparse_ir::tsvd,
sparse_ir::sve::compute_sve_dense, …), where they follow the semver
of those crates; the crates of the workspace require each other at the
same exact version, so such a change reaches them in a release of their
own.
Modules§
- fpu_
check - FPU state checking for Intel Fortran compatibility
- kernelmatrix
- Kernel matrices on the Gauss points of the SVE
- parallel
- Parallel execution on the rayon pool
- sve
- The strategies
compute_svedecomposes a kernel with, and the dense reference path - tsvd
- Truncated SVD in
f64and double-double, by nalgebra
Functions§
- require_
accuracy Okfor an accuracy that selects the SVE:None(automatic) or a value in (0, 1)- require_
finite Okifvalueis finite- require_
nonzero_ size OkunlesssizeisSome(0)- require_
positive_ finite Okifvalueis positive and finite- require_
threshold Okfor a relative truncation threshold:Noneor a value in [0, 1), where 0 keeps every singular value