CryptOpt: Automatic Optimization of Straightline Code

Joel Kuepper, David Wu, Andres Erbsen, Jason Gross, Owen Conoly, Chuyue Sun, Samuel Tian, Adam Chlipala, Chitchanok Chuengsatiansup, Daniel Genkin, Markus Wagner, Yuval Yarom

Abstract: Manual engineering of high-performance implemen-tations typically consumes many resources and requires in-depth knowledge of the hardware. Compilers try to address these problems; however, they are limited by design in what they can do. To address this, we present CryptOpt, an automatic optimizer for long stretches of straightline code. Experimental results across eight hardware platforms show that CryptOpt achieves a speed-up factor of up to 2.56 over current off-the-shelf compilers.