Automated GUI testing plays a crucial role for smartphone vendors who have to ensure that the widely used mobile apps-that are not essentially developed by the vendors-are compatible with new devices and system updates. While existing testing techniques can automatically generate event sequences to reach different GUI views, inputs such as strings and numbers remain difficult to generate, as their generation often involves semantic understanding of the app functionality. Recently, Large Language Models (LLMs) have been successfully adopted to generate string inputs that are semantically relevant to the test case. This paper evaluates the LLM-based input generation in the industrial context of vendor testing of both in-house and 3rd party mobile apps. We present DROIDFILLER, an LLM based input generation technique that builds upon existing work with more sophisticated prompt engineering and customisable context retrieval. DROIDFILLER is empirically evaluated using a total of 120 textfields collected from a total of 45 apps, including both in-house and 3rd party ones. The results show that DROIDFILLER can outperform both vanilla LLM based input generation as well as the existing resource pool approach. We integrate DROIDFILLER into the existing GUI testing framework used at Samsung, evaluate its performance, and discuss the challenges and considerations for practical adoption of LLM-based input generation in the industry.

Emerging real-time multi-model ML (RTMM) workloads such as AR/VR and drone control involve dynamic behaviors in various granularity; task, model, and layers within a model. Such dynamic behaviors introduce new challenges to the system software in an ML system since the overall system load is not completely predictable, unlike traditional ML workloads. In addition, RTMM workloads require real-time processing, involve highly heterogeneous models, and target resource-constrained devices. Under such circumstances, developing an effective scheduler gains more importance to better utilize underlying hardware considering the unique characteristics of RTMM workloads. Therefore, we propose a new scheduler, DREAM, which effectively handles various dynamicity in RTMM workloads targeting multi-accelerator systems. DREAM quantifies the unique requirements for RTMM workloads and utilizes the quantified scores to drive scheduling decisions, considering the current system load and other inference jobs on different models and input frames. DREAM utilizes tunable parameters that provide fast and effective adaptivity to dynamic workload changes. In our evaluation of five scenarios of RTMM workload, DREAM reduces the overall UXCosT, which is an equivalent metric of the energy-delay product (EDP) for RTMM defined in the paper, by 32.2% and 50.0% in the geometric mean (up to 80.8% and 97.6%) compared to state-of-the-art baselines, which shows the efficacy of our scheduling methodology.

Driven by the wide adoption of deep neural networks (DNNs) across different application domains, multi-tenancy execution, where multiple DNNs are deployed simultaneously on the same hardware, has been proposed to satisfy the latency requirements of different applications while improving the overall system utilization. However, multi-tenancy execution could lead to undesired system-level resource contention, causing quality-of-service (QoS) degradation for latency-critical applications.To address this challenge, we propose MoCA1, an adaptive multi-tenancy system for DNN accelerators. Unlike existing solutions that focus on compute resource partition, MoCA dynamically manages shared memory resources of co-located applications to meet their QoS targets. Specifically, MoCA leverages the regularities in both DNN operators and accelerators to dynamically modulate memory access rates based on their latency targets and user-defined priorities so that co-located applications get the resources they demand without significantly starving their co-runners. We demonstrate that MoCA improves the satisfaction rate of the service level agreement (SLA) up to 3.9× (1.8× average), system throughput by 2.3× (1.7× average), and fairness by 1.3× (1.2× average), compared to prior work.

Robotic systems, such as autonomous unmanned aerial vehicles (UAVs) and self-driving cars, have been widely deployed in many scenarios and have the potential to revolutionize the future generation of computing. To improve the performance and energy efficiency of robotic platforms, significant research efforts are being devoted to developing hardware accelerators for workloads that form bottlenecks in the robotics software pipeline. Although domain-specific accelerators can offer improved efficiency over general-purpose processors on isolated robotics benchmarks, system-level constraints such as data movement and contention over shared resources can significantly impact the achievable end-to-end acceleration. In addition, the closed-loop nature of robotic systems, where there is a tight interaction across different deployed environments, software stacks, and hardware architecture, further exacerbates the difficulties of evaluating robotics SoCs.

With the widespread adoption of deep neural networks (DNNs) across applications, there is a growing demand for DNN deployment solutions that can seamlessly support multi-tenant execution. This involves simultaneously running multiple DNN workloads on heterogeneous architectures with domain-specific accelerators. However, existing accelerator interfaces directly bind the accelerator’s physical resources to user threads, without an efficient mechanism to adaptively re-partition available resources. This leads to high programming complexities and performance overheads due to sub-optimal resource allocation, making scalable many-accelerator deployment impractical.

As Deep Neural Networks (DNNs) are rapidly adopted in various domains, many test adequacy metrics for DNN inputs have been introduced to help evaluating, and validating, trained DNN models. Surprise Adequacy (SA) is one such metric that aims to quantitatively measure how surprising a new input is with respect to the data used to train the given model. While SA has been shown to be effective for computer vision tasks such as image classification or object segmentation, its efficacy for DNN based Natural Language Processing has not been thoroughly studied. This paper evaluates whether it is feasible to apply SA analysis to DNN models trained for NLP tasks. We also show that the input distribution captured in the latent embedding space can be multimodal1for some NLP tasks, unlike those observed in computer vision tasks, and investigate if catering for the multimodal property of NLP models can improve SA analysis. An empirical evaluation of extended SA metrics with three NLP tasks and nine DNN models shows that, while unimodal SAs perform sufficiently well for text classification, multimodal SA can outperform unimodal metrics.

DNN accelerators are often developed and evaluated in isolation without considering the cross-stack, system-level effects in real-world environments. This makes it difficult to appreciate the impact of Systemon-Chip (SoC) resource contention, OS overheads, and programming-stack inefficiencies on overall performance/energy-efficiency. To address this challenge, we present Gemmini, an open-source, full-stack DNN accelerator generator. Gemmini generates a wide design-space of efficient ASIC accelerators from a flexible architectural template, together with flexible programming stacks and full SoCs with shared resources that capture system-level effects. Gemmini-generated accelerators have also been fabricated, delivering up to three orders-of-magnitude speedups over high-performance CPUs on various DNN benchmarks.