Dedicated tensor accelerators demonstrate the importance of linear algebra in modern applications. Such accelerators have the potential for impressive performance gains, but require programmers to rewrite code using vendor APIs - a barrier to wider scale adoption. Recent work overcomes this by matching and replacing patterns within code, but such approaches are fragile and fail to cope with the diversity of real-world codes. We develop ATC, a compiler that uses program synthesis to map regions of code to specific APIs. The mapping space that ATC explores is combinatorially large, requiring the development of program classification, dynamic analysis, variable constraint generation and lexical distance matching techniques to make it tractable. We apply ATC to real-world tensor and linear algebra codes and evaluate them against four state-of-the-art approaches. We accelerate between 2.6x and 7x more programs, leading to over an order of magnitude performance improvement.
相關內容
During the past decades, evolutionary computation (EC) has demonstrated promising potential in solving various complex optimization problems of relatively small scales. Nowadays, however, ongoing developments in modern science and engineering are bringing increasingly grave challenges to the conventional EC paradigm in terms of scalability. As problem scales increase, on the one hand, the encoding spaces (i.e., dimensions of the decision vectors) are intrinsically larger; on the other hand, EC algorithms often require growing numbers of function evaluations (and probably larger population sizes as well) to work properly. To meet such emerging challenges, not only does it require delicate algorithm designs, but more importantly, a high-performance computing framework is indispensable. Hence, we develop a distributed GPU-accelerated algorithm library -- EvoX. First, we propose a generalized workflow for implementing general EC algorithms. Second, we design a scalable computing framework for running EC algorithms on distributed GPU devices. Third, we provide user-friendly interfaces to both researchers and practitioners for benchmark studies as well as extended real-world applications. To comprehensively assess the performance of EvoX, we conduct a series of experiments, including: (i) scalability test via numerical optimization benchmarks with problem dimensions/population sizes up to millions; (ii) acceleration test via a neuroevolution task with multiple GPU nodes; (iii) extensibility demonstration via the application to reinforcement learning tasks on the OpenAI Gym. The code of EvoX is available at //github.com/EMI-Group/EvoX.
This paper studies a house allocation problem in a networked housing market, where agents can invite others to join the system in order to enrich their options. Top Trading Cycle is a well-known matching mechanism that achieves a set of desirable properties in a market without invitations. However, under a tree-structured networked market, existing agents have to strategically propagate the barter market as their invitees may compete in the same house with them. Our impossibility result shows that TTC cannot work properly in a networked housing market. Hence, we characterize the possible competitions between inviters and invitees, which lead agents to fail to refer others truthfully (strategy-proof). We then present a novel mechanism based on TTC, avoiding the aforementioned competition to ensure all agents report preference and propagate the barter market truthfully. Unlike the existing mechanisms, the agents' preferences are less restricted under our mechanism. Furthermore, we show by simulations that our mechanism outperforms the existing matching mechanisms in terms of the number of swaps and agents' satisfaction.
Value iteration can find the optimal replenishment policy for a perishable inventory problem, but is computationally demanding due to the large state spaces that are required to represent the age profile of stock. The parallel processing capabilities of modern GPUs can reduce the wall time required to run value iteration by updating many states simultaneously. The adoption of GPU-accelerated approaches has been limited in operational research relative to other fields like machine learning, in which new software frameworks have made GPU programming widely accessible. We used the Python library JAX to implement value iteration and simulators of the underlying Markov decision processes in a high-level API, and relied on this library's function transformations and compiler to efficiently utilize GPU hardware. Our method can extend use of value iteration to settings that were previously considered infeasible or impractical. We demonstrate this on example scenarios from three recent studies which include problems with over 16 million states and additional problem features, such as substitution between products, that increase computational complexity. We compare the performance of the optimal replenishment policies to heuristic policies, fitted using simulation optimization in JAX which allowed the parallel evaluation of multiple candidate policy parameters on thousands of simulated years. The heuristic policies gave a maximum optimality gap of 2.49%. Our general approach may be applicable to a wide range of problems in operational research that would benefit from large-scale parallel computation on consumer-grade GPU hardware.
State-of-art NPUs are typically architected as a self-contained sub-system with multiple heterogeneous hardware computing modules, and a dataflow-driven programming model. There lacks well-established methodology and tools in the industry to evaluate and compare the performance of NPUs from different architectures. We present an event-based performance modeling framework, VPU-EM, targeting scalable performance evaluation of modern NPUs across diversified AI workloads. The framework adopts high-level event-based system-simulation methodology to abstract away design details for speed, while maintaining hardware pipelining, concurrency and interaction with software task scheduling. It is natively developed in Python and built to interface directly with AI frameworks such as Tensorflow, PyTorch, ONNX and OpenVINO, linking various in-house NPU graph compilers to achieve optimized full model performance. Furthermore, VPU-EM also provides the capability to model power characteristics of NPU in Power-EM mode to enable joint performance/power analysis. Using VPU-EM, we conduct performance/power analysis of models from representative neural network architecture. We demonstrate that even though this framework is developed for Intel VPU, an Intel in-house NPU IP technology, the methodology can be generalized for analysis of modern NPUs.
Automated program repair (APR) faces the challenge of test overfitting, where generated patches pass validation tests but fail to generalize. Existing methods for patch assessment involve generating new tests or manual inspection, which can be time-consuming or biased. In this paper, we propose a novel technique, INVALIDATOR, to automatically assess the correctness of APR-generated patches via semantic and syntactic reasoning. INVALIDATOR leverages program invariants to reason about program semantics while also capturing program syntax through language semantics learned from a large code corpus using a pre-trained language model. Given a buggy program and the developer-patched program, INVALIDATOR infers likely invariants on both programs. Then, INVALIDATOR determines that an APR-generated patch overfits if: (1) it violates correct specifications or (2) maintains erroneous behaviors from the original buggy program. In case our approach fails to determine an overfitting patch based on invariants, INVALIDATOR utilizes a trained model from labeled patches to assess patch correctness based on program syntax. The benefit of INVALIDATOR is threefold. First, INVALIDATOR leverages both semantic and syntactic reasoning to enhance its discriminative capability. Second, INVALIDATOR does not require new test cases to be generated, but instead only relies on the current test suite and uses invariant inference to generalize program behaviors. Third, INVALIDATOR is fully automated. Experimental results demonstrate that INVALIDATOR outperforms existing methods in terms of Accuracy and F-measure, correctly identifying 79% of overfitting patches and detecting 23% more overfitting patches than the best baseline.
Jupyter notebooks enable developers to interleave code snippets with rich-text and in-line visualizations. Data scientists use Jupyter notebook as the de-facto standard for creating and sharing machine-learning based solutions, primarily written in Python. Recent studies have demonstrated, however, that a large portion of Jupyter notebooks available on public platforms are undocumented and lacks a narrative structure. This reduces the readability of these notebooks. To address this shortcoming, this paper presents HeaderGen, a novel tool-based approach that automatically annotates code cells with categorical markdown headers based on a taxonomy of machine-learning operations, and classifies and displays function calls according to this taxonomy. For this functionality to be realized, HeaderGen enhances an existing call graph analysis in PyCG. To improve precision, HeaderGen extends PyCG's analysis with support for handling external library code and flow-sensitivity. The former is realized by facilitating the resolution of function return-types. Furthermore, HeaderGen uses type information to perform pattern matching on code syntax to annotate code cells. The evaluation on 15 real-world Jupyter notebooks from Kaggle shows that HeaderGen's underlying call graph analysis yields high accuracy (96.4% precision and 95.9% recall). This is because HeaderGen can resolve return-types of external libraries where existing type inference tools such as pytype (by Google), pyright (by Microsoft), and Jedi fall short. The header generation has a precision of 82.2% and a recall rate of 96.8% with regard to headers created manually by experts. In a user study, HeaderGen helps participants finish comprehension and navigation tasks faster. All participants clearly perceive HeaderGen as useful to their task.
Passwords are the most common mechanism for authenticating users online. However, studies have shown that users find it difficult to create and manage secure passwords. To that end, passphrases are often recommended as a usable alternative to passwords, which would potentially be easy to remember and hard to guess. However, as we show, user-chosen passphrases fall short of being secure, while state-of-the-art machine-generated passphrases are difficult to remember. In this work, we aim to tackle the drawbacks of the systems that generate passphrases for practical use. In particular, we address the problem of generating secure and memorable passphrases and compare them against user chosen passphrases in use. We identify and characterize 72, 999 user-chosen in-use unique English passphrases from prior leaked password databases. Then we leverage this understanding to create a novel framework for measuring memorability and guessability of passphrases. Utilizing our framework, we design MASCARA, which follows a constrained Markov generation process to create passphrases that optimize for both memorability and guessability. Our evaluation of passphrases shows that MASCARA-generated passphrases are harder to guess than in-use user-generated passphrases, while being easier to remember compared to state-of-the-art machine-generated passphrases. We conduct a two-part user study with crowdsourcing platform Prolific to demonstrate that users have highest memory-recall (and lowest error rate) while using MASCARA passphrases. Moreover, for passphrases of length desired by the users, the recall rate is 60-100% higher for MASCARA-generated passphrases compared to current system-generated ones.
Recent years have witnessed impressive progress in super-resolution (SR) processing. However, its real-time inference requirement sets a challenge not only for the model design but also for the on-chip implementation. In this paper, we implement a full-stack SR acceleration framework on embedded GPU devices. The special dictionary learning algorithm used in SR models was analyzed in detail and accelerated via a novel dictionary selective strategy. Besides, the hardware programming architecture together with the model structure is analyzed to guide the optimal design of computation kernels to minimize the inference latency under the resource constraints. With these novel techniques, the communication and computation bottlenecks in the deep dictionary learning-based SR models are tackled perfectly. The experiments on the edge embedded NVIDIA NX and 2080Ti show that our method outperforms the state-of-the-art NVIDIA TensorRT significantly, and can achieve real-time performance.
This paper presents a parallel preconditioning approach based on incomplete LU (ILU) factorizations in the framework of Domain Decomposition (DD) for general sparse linear systems. We focus on distributed memory parallel architectures, specifically, those that are equipped with graphic processing units (GPUs). In addition to block Jacobi, we present general purpose two-level ILU Schur complement-based approaches, where different strategies are presented to solve the coarse-level reduced system. These strategies are combined with modified ILU methods in the construction of the coarse-level operator, in order to effectively remove smooth errors. We leverage available GPU-based sparse matrix kernels to accelerate the setup and the solve phases of the proposed ILU preconditioner. We evaluate the efficiency of the proposed methods as a smoother for algebraic multigrid (AMG) and as a preconditioner for Krylov subspace methods, on challenging anisotropic diffusion problems and a collection of general sparse matrices.
As online dating has become more popular in the past few years, an efficient and effective algorithm to match users is needed. In this project, we proposed a new dating matching algorithm that uses Kendall-Tau distance to measure the similarity between users based on their ranking for items in a list. (e.g., their favourite sports, music, etc.) To increase the performance of the search process, we applied a tree-based searching structure, Cascading Metric Tree (CMT), on this metric. The tree is built on ranked lists from all the users; when a query target and a radius are provided, our algorithm can return users within the radius of the target. We tested the scaling of this searching method on a synthetic dataset by varying list length, population size, and query radius. We observed that the algorithm is able to query the best matching people for the user in a practical time, given reasonable parameters. We also provided potential future improvements that can be made to this algorithm based on the limitations. Finally, we offered more use cases of this search structure on Kendall-Tau distance and new insight into real-world applications of distance search structures.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191