dense
DenseVectorKernels supplies Level 1 arithmetic over array windows and strides. DenseBlas supplies the Level 2 and 3 contracts, with shared portable traversal, validation, no-read behavior and alias staging.
Shared internal kernels supply panel arithmetic, packed product tiles and diagonal substitutions. Transient packing and coefficient gathering borrow scoped scratch from com.eignex.koblas.Workspace.
Dense matrix operands use com.eignex.koblas.DenseMatrixRegion and vectors use com.eignex.koblas.DenseVector, preserving offsets, leading dimensions and signed strides; transposition and MatrixStructure travel as flags. gemmt is the triangle-selected general product Netlib calls GEMMTR: full column-major operands and a full-storage selected output triangle, not compact BLAS packed storage. A host binding without cblas_dgemmt composes the operation from gemm.
Types
Contiguous dense Level 1 operations that are independently useful outside matrix algorithms. Implementations must read and write only the array windows selected by their offsets and lengths. Callers must supply non-negative lengths and valid windows; implementations need not validate them. A zero length is legal for every operation that accepts one and produces the documented empty result without reading either array.
Structure of a stored matrix; unstored entries and unit diagonals are never loaded.