GauS: Differentiable Scheduling Optimization via Gaussian Reparameterization
This addresses a fundamental problem in compilation and synthesis for developers and engineers, offering a novel approach to replace traditional methods.
The paper tackled the challenge of efficient operator scheduling in software compilation and hardware synthesis by proposing GauS, a differentiable framework that models scheduling with Gaussian distributions, achieving Pareto-optimal results on benchmarks.
Efficient operator scheduling is a fundamental challenge in software compilation and hardware synthesis. While recent differentiable approaches have sought to replace traditional ones like exact solvers or heuristics with gradient-based search, they typically rely on categorical distributions that fail to capture the ordinal nature of time and suffer from a parameter space that scales poorly. In this paper, we propose a novel differentiable framework, GauS, that models operator scheduling as a stochastic relaxation using Gaussian distributions, which fully utilize modern parallel computing devices like GPUs. By representing schedules as continuous Gaussian variables, we successfully capture the ordinal nature of time and reduce the optimization space by orders of magnitude. Our method is highly flexible to represent various objectives and constraints, which provides the first differentiable formulation for the complex pipelined scheduling problem. We evaluate our method on a range of benchmarks, demonstrating that Gaus achieves Pareto-optimal results.