Module block_tensor 

Block tensor type for GMRES with block matrices.

This module provides BlockTensor, a collection of tensors organized in a block structure. It implements TensorLike for the vector space operations required by GMRES, allowing block matrix linear equations Ax = b to be solved using the existing GMRES implementation.

Example

use tensor4all_core::block_tensor::BlockTensor;
use tensor4all_core::krylov::{gmres, GmresOptions};
use tensor4all_core::{DynIndex, TensorDynLen};

let i = DynIndex::new_dyn(2);
let b_block = TensorDynLen::from_dense(vec![i.clone()], vec![1.0, 2.0])?;
let zero_block = TensorDynLen::from_dense(vec![i.clone()], vec![0.0, 0.0])?;

let b = BlockTensor::new(vec![b_block], (1, 1));
let x0 = BlockTensor::new(vec![zero_block], (1, 1));

let apply_a = |x: &BlockTensor<TensorDynLen>| Ok(x.clone());
let result = gmres(apply_a, &b, &x0, &GmresOptions::default())?;

assert!(result.converged);
assert_eq!(result.solution.shape(), (1, 1));

Structs

BlockTensor

A collection of tensors organized in a block structure.