An LP-based methodology for improved timing-driven placement