Back-transformation algorithm (QR-like)

This set of scripts has been used to understand back-transformation algorithm (with reflectors applied from 0 to k, with results comparable to a QR decomposition: Q_n Q_n-1 ... Q₁ C).

Creating a random matrix

python createTestCase.py -n <n>

This script creates a n x n matrix, filled with integers ranging from 0 to 30.

Compute Reflectors and Taus

python computeReflectorsAndTaus.py --input matrix.py

Using the output file of the previous step, reflectors and taus are computed with this script.

Computing T factor matrix

python computeTfactor.py --input myReflectorsAndTaus.py

This script takes as input the previously computed reflectors and taus and compute T factor matrix.

Collect data

bash collect4backtrans.sh

This bash script collect starting matrix (A), matrix of reflectors (V) and matrix of T factor (T).

Back-transformation

This first python script applies the back-transformation using matrix multiplications (solving A = A - V T V* A):

python matmul_backtrans.py --set file test-backtrans.py

This second python script applies the algorithm tuned for single element (or tiles in DLAF):

python analytical_backtrans.py --set file test-backtrans.py