A technique that partitions large matrices into smaller blocks and processes them independently to improve computational efficiency.