In this paper, we propose a novel and unified dynamic intelligent reflecting surface (IRS) beamforming framework to enhance the sum computation rate of an mobile edge computing (MEC) system, where each device follows a binary offloading policy. Our goal is to maximize the sum computational bits by jointly optimizing the computational mode of each device, IRS beamforming vectors across time, and offloading time allocation. Since the resulting optimization problem is non-convex, there are generally no standard methods to solve it optimally. To tackle this problem, we first derive the offloading activation condition for each device by deeply exploiting the intrinsic structure of the original optimization problem. Based on the offloading activation condition, we proposed an efficient algorithm to obtain the high-quality solution of the computation rate maximization problem. The optimal condition of the proposed algorithm is also revealed. Numerical results demonstrate the superiority of our proposed design over benchmark schemes and also unveil the fundamental performance-cost tradeoff of our proposed dynamic IRS beamforming protocol.