Jeremy Bruestle

Evangelos Georganas, Dhiraj Kalamkar, Sasikanth Avancha et al.

During the past decade, novel Deep Learning (DL) algorithms, workloads and hardware have been developed to tackle a wide range of problems. Despite the advances in workload and hardware ecosystems, the programming methodology of DL systems is stagnant. DL workloads leverage either highly-optimized, yet platform-specific and inflexible kernels from DL libraries, or in the case of novel operators, reference implementations are built via DL framework primitives with underwhelming performance. This work introduces the Tensor Processing Primitives (TPP), a programming abstraction striving for efficient, portable implementation of DL workloads with high-productivity. TPPs define a compact, yet versatile set of 2D-tensor operators (or a virtual Tensor ISA), which subsequently can be utilized as building-blocks to construct complex operators on high-dimensional tensors. The TPP specification is platform-agnostic, thus code expressed via TPPs is portable, whereas the TPP implementation is highly-optimized and platform-specific. We demonstrate the efficacy and viability of our approach using standalone kernels and end-to-end DL & HPC workloads expressed entirely via TPPs that outperform state-of-the-art implementations on multiple platforms.

Tim Zerrell, Jeremy Bruestle

Hardware architectures and machine learning (ML) libraries evolve rapidly. Traditional compilers often fail to generate high-performance code across the spectrum of new hardware offerings. To mitigate, engineers develop hand-tuned kernels for each ML library update and hardware upgrade. Unfortunately, this approach requires excessive engineering effort to scale or maintain with any degree of state-of-the-art performance. Here we present a Nested Polyhedral Model for representing highly parallelizable computations with limited dependencies between iterations. This model provides an underlying framework for an intermediate representation (IR) called Stripe, amenable to standard compiler techniques while naturally modeling key aspects of modern ML computing. Stripe represents parallelism, efficient memory layout, and multiple compute units at a level of abstraction amenable to automatic optimization. We describe how Stripe enables a compiler for ML in the style of LLVM that allows independent development of algorithms, optimizations, and hardware accelerators. We also discuss the design exploration advantages of Stripe over kernel libraries and schedule-based or schedule-space-based code generation.