C++ matrix API overview¶

tiny::Mat uses float, row-major storage and optional row padding. Owned buffers are released at destruction; callers must keep external buffers and ROI parent storage alive.

Reading goal	Page
Construction, copying, external buffers and layout	Objects and memory
ROI, data manipulation, arithmetic and streams	Operations
Determinants, inverses and linear systems	Linear systems
LU, Cholesky, QR, SVD and solving	Decompositions
Power iteration, Jacobi and QR	Eigenvalues
Full header and original links	Complete reference
Implementation and verification	Source · Test overview

Minimal usage¶

#include "tiny_matrix.hpp"

void matrix_demo()
{
    tiny::Mat A(2, 2);
    if (!A.data) return;
    A(0, 0) = 4.0f; A(0, 1) = 1.0f;
    A(1, 0) = 1.0f; A(1, 1) = 3.0f;
    auto lu = A.lu_decompose();
    if (lu.status != TINY_OK) return;
    lu.L.print_matrix(false);
    lu.U.print_matrix(false);
}

Usage boundaries¶

Fields are row/col/step; stride is a legacy alias of step in this class. Elements live at data[i * step + j].
C tiny_mat padding/stride arguments are not interchangeable with C++ metadata.
Preconditions and failure states depend on the method. The QR eigenvalue API returns only real parts of complex results.
Timing, code size and stability require measurements for the actual input and build; unspecified historical speedup factors are not guarantees.