|
      
|
|
|
|
Iterative compilers have been shown to produce good results for programs designed to run on embedded systems, at the cost of very long compilation time. This talk focuses on techniques such as probabilistic methods, Markov models and unsupervised methods to both dramatically reduce the time needed for search, and to ensure that time spent learning is most effectively utilised, looking to make learning compilers viable for general purpose compilation.
