Obtaining high-performance implementations of tensor programs such as deep neural networks on a wide range of hardware remains a challenging task. Search-based tensor program optimizers can automatically find high-performance programs on a given hardware platform, but the search process in existing tools suffer from low efficiency, requiring hours or days of time to discover good programs due to the size of the search space.

Runtime-tunable context-dependent network compression would make mobile deep learning (DL) adaptable to often varying resource availability, input “difficulty”, or user needs. The existing compression techniques significantly reduce the memory, processing, and energy tax of DL, yet, the resulting models tend to be permanently impaired, sacrificing the inference power for reduced resource usage. The existing tunable compression approaches, on the other hand, require expensive re-training, do not support arbitrary strategies for adapting the compression and do not provide mobile-ready implementations. In this paper we present Mobiprox, a framework enabling mobile DL with flexible precision. Mobiprox implements tunable approximations of tensor operations and enables runtime-adaptable approximation of the individual network layers. A profiler and a tuner included with Mobiprox identify the most promising neural network approximation configurations leading to the desired inference quality with the minimal use of resources. Furthermore, we develop control strategies that depending on contextual factors, such as the input data difficulty, dynamically adjust the approximation levels across a mobile DL model’s layers. We implement Mobiprox in Android OS and through experiments in diverse mobile domains, including human activity recognition and spoken keyword detection, demonstrate that it can save up to 15% system-wide energy with a minimal impact on the inference accuracy.

Extended reality (XR) devices, including augmented, virtual, and mixed reality, provide a deeply immersive experience. However, practical limitations like weight, heat, and comfort put extreme constraints on the performance, power consumption, and image quality of such systems. In this paper, we study how these constraints form the tradeoff between Fixed Foveated Rendering (FFR), Gaze-Tracked Foveated Rendering (TFR), and conventional, non-foveated rendering. While existing papers have often studied these methods, we provide the first comprehensive study of their relative feasibility in practical systems with limited battery life and computational budget. We show that TFR with the added cost of the gaze-tracker can often be more expensive than FFR. Thus, we co-design a gaze-tracked foveated renderer considering its benefits in computation, power efficiency, and tradeoffs in image quality. We describe principled approximations for eye tracking which provide up to a 9x speedup in runtime performance with approximately a 20x improvement in energy efficiency when run on a mobile GPU. In isolation, these approximations appear to significantly degrade the gaze quality, but appropriate compensation in the visual pipeline can mitigate the loss. Overall, we show that with a highly optimized gaze-tracker, TFR is feasible compared to FFR, resulting in up to 1.25x faster frame times while also reducing total energy consumption by over 40%.

We present Trimmer, a state-of-the-art tool for reducing code size. Trimmer reduces code sizes by specializing programs with respect to constant inputs provided by developers. The static data can be provided as command-line options or through configuration files. The constants define the features that must be retained, which in turn determine the features that are unused in a specific deployment (and can therefore be removed). Trimmer includes sophisticated compiler transformations for input specialization, supports precise yet efficient context-sensitive inter-procedural constant propagation, and introduces a custom loop unroller. Trimmer is easy-to-use and extensively parameterized. We discuss how Trimmer can be configured by developers to explicitly trade analysis precision and specialization time. We also provide a high-level description of Trimmer’s static analysis passes. The source code is publicly available at: https://github.com/ashish-gehani/Trimmer. A video demonstration can be found here: https://youtu.be/6pAuJ68INnI.

Software bloat has negative implications for security, reliability, and performance. To counter bloat, we proposeTrimmer, a static analysis-based system for pruning unused functionality.Trimmerremoves code that is unused with respect to user-provided command-line arguments and application-specific configuration files.Trimmeruses concrete memory tracking and a custom inter-procedural constant propagation analysis that facilitates dead code elimination. Our system supports both context-sensitive and context-insensitive constant propagation. We show that context-sensitive constant propagation is important for effective software pruning in most applications. We introducesparse constant propagationthat performs constant propagation only for configuration-hosting variables and show that it performs better (higher code size reductions) compared to constant propagation for all program variables. Overall, our results show thatTrimmerreduces binary sizes for real-world programs with reasonable analysis times. Across 20 evaluated programs, we observe a mean binary size reduction of 22.7 percent and a maximum reduction of 62.7 percent. For 5 programs, we observe performance speedups ranging from 5 to 53 percent. Moreover, we show that winnowing software applications can reduce the program attack surface by removing code that contains exploitable vulnerabilities. We find that debloating usingTrimmerremoves CVEs in 4 applications.

Manually optimizing the tradeoffs between accuracy, performance and energy for resource-intensive applications with flexible accuracy or precision requirements is extremely difficult. We present ApproxTuner, an automatic framework for accuracy-aware optimization of tensor-based applications while requiring only high-level end-to-end quality specifications. ApproxTuner implements and manages approximations in algorithms, system software, and hardware.

We propose ApproxHPVM, a compiler IR and system designed to enable accuracy-aware performance and energy tuning on heterogeneous systems with multiple compute units and approximation methods. ApproxHPVM automatically translates end-to-end application-level quality metrics into accuracy requirements for individual operations. ApproxHPVM uses a hardware-agnostic accuracy-tuning phase to do this translation that provides greater portability across heterogeneous hardware platforms and enables future capabilities like accuracy-aware dynamic scheduling and design space exploration. ApproxHPVM incorporates three main components: (a) a compiler IR with hardware-agnostic approximation metrics, (b) a hardware-agnostic accuracy-tuning phase to identify error-tolerant computations, and (c) an accuracy-aware hardware scheduler that maps error-tolerant computations to approximate hardware components. As ApproxHPVM does not incorporate any hardware-specific knowledge as part of the IR, it can serve as a portable virtual ISA that can be shipped to all kinds of hardware platforms. We evaluate our framework on nine benchmarks from the deep learning domain and five image processing benchmarks. Our results show that our framework can offload chunks of approximable computations to special-purpose accelerators that provide significant gains in performance and energy, while staying within user-specified application-level quality metrics with high probability. Across the 14 benchmarks, we observe from 1-9x performance speedups and 1.1-11.3x energy reduction for very small reductions in accuracy.

With the proliferation of new hardware architectures and ever-evolving user requirements, the software stack is becoming increasingly bloated. In practice, only a limited subset of the supported functionality is utilized in a particular usage context, thereby presenting an opportunity to eliminate unused features. In the past, program specialization has been proposed as a mechanism for enabling automatic software debloating. In this work, we show how existing program specialization techniques lack the analyses required for providing code simplification for real-world programs. We present an approach that uses stronger analysis techniques to take advantage of constant configuration data, thereby enabling more effective debloating. We developed Trimmer, an application specialization tool that leverages user-provided configuration data to specialize an application to its deployment context. The specialization process attempts to eliminate the application functionality that is unused in the user-defined context. Our evaluation demonstrates Trimmer can effectively reduce code bloat. For 13 applications spanning various domains, we observe a mean binary size reduction of 21% and a maximum reduction of 75%. We also show specialization reduces the surface for code-reuse attacks by reducing the number of exploitable gadgets. For the evaluated programs, we observe a 20% mean reduction in the total gadget count and a maximum reduction of 87%.