1. A new approach to approximate computing is proposed, where programmers use a type system to communicate high-level constraints on the degree of approximation.
2. The core type system captures the probability that each operation exhibits an error and bounds the probability that each expression deviates from its correct value.
3. Solver-aided type inference lets the programmer specify the correctness probability on only some variables and automatically fills in other types to meet these specifications.
The article presents a new approach for approximate computing, which uses a type system to communicate high-level constraints on the degree of approximation. The core type system captures the probability that each operation exhibits an error and bounds the probability that each expression deviates from its correct value. Solver-aided type inference allows programmers to specify correctness probabilities on certain variables while automatically filling in other types to meet these specifications.
The article is generally reliable and trustworthy, as it provides evidence for its claims through existing approximate-computing benchmarks and examines implications for approximate hardware design. It also offers a strong soundness guarantee by using solver-aided optimization to improve efficiency. However, there are some potential biases in the article, such as one-sided reporting or partiality towards certain approaches or technologies over others. Additionally, there may be missing points of consideration or unexplored counterarguments that could have been addressed in order to provide a more comprehensive overview of the topic at hand.