Fully Homomorphic Encryption (FHE) is a cryptographic method that guarantees the privacy and security of user data during computation. FHE algorithms can perform unlimited arithmetic computations directly on encrypted data without decrypting it. Thus, even when processed by untrusted systems, confidential data is never exposed. In this work, we develop new techniques for accelerated encrypted execution and demonstrate the significant performance advantages of our approach. Our current focus is the Fully Homomorphic Encryption over the Torus (CGGI) scheme, which is a current state-of-the-art method for evaluating arbitrary functions in the encrypted domain. CGGI represents a computation as a graph of homomorphic logic gates and each individual bit of the plaintext is transformed into a polynomial in the encrypted domain. Arithmetic on such data becomes very expensive: operations on bits become operations on entire polynomials. Therefore, evaluating even relatively simple nonlinear functions with the CGGI cryptosystem, such as a sigmoid, can take thousands of seconds on a single CPU thread. Using our novel framework for end-to-end accelerated encrypted execution called ArctyrEX, developers with no knowledge of complex FHE libraries can simply describe their computation as a C program that is evaluated 18x faster on average relative to the GPU-accelerated Concrete library for multiplication-intensive benchmarks.

Tree Search (TS) is crucial to some of the most influential successes in reinforcement learning. Here, we tackle two major challenges with TS that limit its usability: \textit{distribution shift} and \textit{scalability}. We first discover and analyze a counter-intuitive phenomenon: action selection through TS and a pre-trained value function often leads to lower performance compared to the original pre-trained agent, even when having access to the exact state and reward in future steps. We show this is due to a distribution shift to areas where value estimates are highly inaccurate and analyze this effect using Extreme Value theory. To overcome this problem, we introduce a novel off-policy correction term that accounts for the mismatch between the pre-trained value and its corresponding TS policy by penalizing under-sampled trajectories. We prove that our correction eliminates the above mismatch and bound the probability of sub-optimal action selection. Our correction significantly improves pre-trained Rainbow agents without any further training, often more than doubling their scores on Atari games. Next, we address the scalability issue given by the computational complexity of exhaustive TS that scales exponentially with the tree depth. We introduce Batch-BFS: a GPU breadth-first search that advances all nodes in each depth of the tree simultaneously. Batch-BFS reduces runtime by two orders of magnitude and, beyond inference, enables also training with TS of depths that were not feasible before. We train DQN agents from scratch using TS and show improvement in several Atari games compared to both the original DQN and the more advanced Rainbow. We will share the code upon publication.

We introduce CuLE (CUDA Learning Environment), a CUDA port of the Atari Learning Environment (ALE) which is used for the development of deep reinforcement algorithms. CuLE overcomes many limitations of existing CPU-based emulators and scales naturally to multiple GPUs. It leverages GPU parallelization to run thousands of games simultaneously and it renders frames directly on the GPU, to avoid the bottleneck arising from the limited CPU-GPU communication bandwidth. CuLE generates up to 155M frames per hour on a single GPU, a finding previously achieved only through a cluster of CPUs. Beyond highlighting the differences between CPU and GPU emulators in the context of reinforcement learning, we show how to leverage the high throughput of CuLE by effective batching of the training data, and show accelerated convergence for A2C+V-trace. CuLE is available at https://github.com/NVlabs/cule.

Irregular computations on large workloads are a necessity in many areas of computational science. Mapping these computations to modern parallel architectures, such as GPUs, is particularly challenging because the performance often depends critically on the choice of data-structure and algorithm. In this paper, we develop a parallel processing scheme, based on Merge Path partitioning, to compute segmented row-wise operations on sparse matrices that exposes parallelism at the granularity of individual nonzero entries. Our decomposition achieves competitive performance across many diverse problems while maintaining predictable behaviour dependent only on the computational work and ameliorates the impact of irregularity. We evaluate the performance of three sparse kernels: Spiv, Spaded and Sperm. We show that our processing scheme for each kernel yields comparable performance to other schemes in many cases and our performance is highly correlated, nearly 1, to the computational work irrespective of the underlying structure of the matrices.

Sparse matrix--matrix multiplication (SpGEMM) is a key operation in numerous areas from information to the physical sciences. Implementing SpGEMM efficiently on throughput-oriented processors, such as the graphics processing unit (GPU), requires the programmer to expose substantial fine-grained parallelism while conserving the limited off-chip memory bandwidth. Balancing these concerns, we decompose the SpGEMM operation into three highly parallel phases: expansion, sorting, and contraction, and introduce a set of complementary bandwidth-saving performance optimizations. Our implementation is fully general and our optimization strategy adaptively processes the SpGEMM workload row-wise to substantially improve performance by decreasing the work complexity and utilizing the memory hierarchy more effectively.