Ultrasonic imaging and reconstruction tools are commonly used to detect, identify and measure defects in different mechanical parts. Due to the complexity of the underlying physics, and due to the ever growing quantity of acquired data, computation time is becoming a limitation to the optimal inspection of a mechanical part. This article presents the performances of several implementations of a computational heavy algorithm, named Total Focusing Method, on both Graphics Processing Units (GPU) and General Purpose Processors (GPP). The scope of this study is narrowed to planar parts tested in immersion for defects. Using algorithmic simplifications and architectural optimizations, the algorithm has been drastically accelerated resulting in memory-bound implementations. On GPU, high performances can be achieved by profiting from GPU long memory transactions and from hand managed memory. Whereas on GPP, computations cost are overrun by memory access resulting in less efficient performances compared to the computing capabilities available. The following study constitutes the first step toward analyzing the target algorithm for diverse hardware in the non-destructive testing environment.