Seminario: “Approximate Computing Techniques on Loops” di Biagio Cosenza

Abstract: Approximate Computing exploits the gap between the accuracy provided by a system and the accuracy required by an application. Many applications are resistant to some amount of error and earlier works in this field have proven that there is potential for significant improvements in terms of execution time or energy consumption if a small amount of error can be accepted. Many existing approximation techniques focus on loops such as loop perforation, which skips iterations for faster, approximated computation. This presentation will introduce our research work on perforation techniques for both high-end and mobile GPUs. We will see how GPU’s local memory can be very useful to improve the way approximate computation is performed, e.g., by improving the quality of approximation with data reconstruction techniques. Finally, we will discuss how perforation can be extended to a more complex scenario: loop nests. Il collegamento alla ISISLab community della piattaforma Discord è disponibile qui o anche utilizzando il seguente QR Code:

Il seminario si terrà sul canale vocale #seminars della piattaforma Discord è disponibile qui o utilizzando il seguente QR Code: