arXiv subjects

David Novo

Publications and source records attributed to David Novo.

11 recordsLinked to original sources

arXiv2026

Uncertainty-triggered wake-up enables energy-efficient, error-resilient edge AI with memristor front ends

Memristor computing offers a route to low-energy edge AI, but device variability, sensitivity to operating conditions, and system-integration challenges can hinder deployment. Here we show that these limitations can be mitigated by using memristor AI not as the final decision maker but as the ultra-low-power, always-on front end of a heterogeneous inference system. We implement this architecture by coupling a fabricated memristor Bayesian machine to a programmable CPU running a higher-power, higher-accuracy software neural network. The memristor front end acts as a probabilistic screener. When it predicts an abnormal event or produces an ambiguous or invalid output, a dedicated hardware wake-up path activates the CPU, which produces the final decision. We validate this architecture on a heartbeat-classification benchmark by interfacing the fabricated Bayesian machine with an FPGA-based wake-up platform and CPU back end. The resulting uncertainty-triggered wake-up system achieves high final classification accuracy under nominal operation and maintains this accuracy even when the memristor front end is degraded by voltage scaling or reduced programming margins, because unreliable outputs are converted into recoverable wake-up events instead of becoming silent errors. Post-layout analysis of an ASIC implementation shows that average energy is governed primarily by wake-up frequency, providing practical design rules for choosing front-end operating points. These results establish uncertainty-triggered wake-up as a strategy for energy-efficient, error-resilient edge AI.

cs.ET↗

arXiv2025

EasyDRAM: An FPGA-based Infrastructure for Fast and Accurate End-to-End Evaluation of Emerging DRAM Techniques

DRAM is a critical component of modern computing systems. Recent works propose numerous techniques (that we call DRAM techniques) to enhance DRAM-based computing systems' throughput, reliability, and computing capabilities (e.g., in-DRAM bulk data copy). Evaluating the system-wide benefits of DRAM techniques is challenging as they often require modifications across multiple layers of the computing stack. Prior works propose FPGA-based platforms for rapid end-to-end evaluation of DRAM techniques on real DRAM chips. Unfortunately, existing platforms fall short in two major aspects: (1) they require deep expertise in hardware description languages, limiting accessibility; and (2) they are not designed to accurately model modern computing systems. We introduce EasyDRAM, an FPGA-based framework for rapid and accurate end-to-end evaluation of DRAM techniques on real DRAM chips. EasyDRAM overcomes the main drawbacks of prior FPGA-based platforms with two key ideas. First, EasyDRAM removes the need for hardware description language expertise by enabling developers to implement DRAM techniques using a high-level language (C++). At runtime, EasyDRAM executes the software-defined memory system design in a programmable memory controller. Second, EasyDRAM tackles a fundamental challenge in accurately modeling modern systems: real processors typically operate at higher clock frequencies than DRAM, a disparity that is difficult to replicate on FPGA platforms. EasyDRAM addresses this challenge by decoupling the processor-DRAM interface and advancing the system state using a novel technique we call time scaling, which faithfully captures the timing behavior of the modeled system. We believe and hope that EasyDRAM will enable innovative ideas in memory system design to rapidly come to fruition. To aid future research EasyDRAM implementation is open sourced at https://github.com/CMU-SAFARI/EasyDRAM.

cs.AR↗

arXiv2025

Addressing memory bandwidth scalability in vector processors for streaming applications

As the size of artificial intelligence and machine learning (AI/ML) models and datasets grows, the memory bandwidth becomes a critical bottleneck. The paper presents a novel extended memory hierarchy that addresses some major memory bandwidth challenges in data-parallel AI/ML applications. While data-parallel architectures like GPUs and neural network accelerators have improved power performance compared to traditional CPUs, they can still be significantly bottlenecked by their memory bandwidth, especially when the data reuse in the loop kernels is limited. Systolic arrays (SAs) and GPUs attempt to mitigate the memory bandwidth bottleneck but can still become memory bandwidth throttled when the amount of data reuse is not sufficient to confine data access mostly to the local memories near to the processing. To mitigate this, the proposed architecture introduces three levels of on-chip memory -- local, intermediate, and global -- with an ultra-wide register and data-shufflers to improve versatility and adaptivity to varying data-parallel applications. The paper explains the innovations at a conceptual level and presents a detailed description of the architecture innovations. We also map a representative data-parallel application, like a convolutional neural network (CNN), to the proposed architecture and quantify the benefits vis-a-vis GPUs and repersentative accelerators based on systolic arrays and vector processors.

cs.AR↗

arXiv2025

DaPPA: A Data-Parallel Programming Framework for Processing-in-Memory Architectures

The growing volume of data in modern applications has led to significant computational costs in conventional processor-centric systems. Processing-in-memory (PIM) architectures alleviate these costs by moving computation closer to memory, reducing data movement overheads. UPMEM is the first commercially available PIM system, featuring thousands of in-order processors (DPUs) integrated within DRAM modules. However, a programming UPMEM-based system remains challenging due to the need for explicit data management and workload partitioning across DPUs. We introduce DaPPA (data-parallel processing-in-memory architecture), a programming framework that eases the programmability of UPMEM systems by automatically managing data movement, memory allocation, and workload distribution. The key idea behind DaPPA is to leverage a high-level data-parallel pattern-based programming interface to abstract hardware complexities away from the programmer. DaPPA comprises three main components: (i) data-parallel pattern APIs, a collection of five primary data-parallel pattern primitives that allow the programmer to express data transformations within an application; (ii) a dataflow programming interface, which allows the programmer to define how data moves across data-parallel patterns; and (iii) a dynamic template-based compilation, which leverages code skeletons and dynamic code transformations to convert data-parallel patterns implemented via the dataflow programming interface into an optimized UPMEM binary. We evaluate DaPPA using six workloads from the PrIM benchmark suite on a real UPMEM system. Compared to hand-tuned implementations, DaPPA improves end-to-end performance by 2.1x, on average, and reduces programming complexity (measured in lines-of-code) by 94%. Our results demonstrate that DaPPA is an effective programming framework for efficient and user-friendly programming on UPMEM systems.

cs.AR↗

arXiv2023

Sibyl: Adaptive and Extensible Data Placement in Hybrid Storage Systems Using Online Reinforcement Learning

Hybrid storage systems (HSS) use multiple different storage devices to provide high and scalable storage capacity at high performance. Recent research proposes various techniques that aim to accurately identify performance-critical data to place it in a "best-fit" storage device. Unfortunately, most of these techniques are rigid, which (1) limits their adaptivity to perform well for a wide range of workloads and storage device configurations, and (2) makes it difficult for designers to extend these techniques to different storage system configurations (e.g., with a different number or different types of storage devices) than the configuration they are designed for. We introduce Sibyl, the first technique that uses reinforcement learning for data placement in hybrid storage systems. Sibyl observes different features of the running workload as well as the storage devices to make system-aware data placement decisions. For every decision it makes, Sibyl receives a reward from the system that it uses to evaluate the long-term performance impact of its decision and continuously optimizes its data placement policy online. We implement Sibyl on real systems with various HSS configurations. Our results show that Sibyl provides 21.6%/19.9% performance improvement in a performance-oriented/cost-oriented HSS configuration compared to the best previous data placement technique. Our evaluation using an HSS configuration with three different storage devices shows that Sibyl outperforms the state-of-the-art data placement policy by 23.9%-48.2%, while significantly reducing the system architect's burden in designing a data placement mechanism that can simultaneously incorporate three storage devices. We show that Sibyl achieves 80% of the performance of an oracle policy that has complete knowledge of future access patterns while incurring a very modest storage overhead of only 124.4 KiB.

cs.AR↗

arXiv2022

Approximations in Deep Learning

The design and implementation of Deep Learning (DL) models is currently receiving a lot of attention from both industrials and academics. However, the computational workload associated with DL is often out of reach for low-power embedded devices and is still costly when run on datacenters. By relaxing the need for fully precise operations, Approximate Computing (AxC) substantially improves performance and energy efficiency. DL is extremely relevant in this context, since playing with the accuracy needed to do adequate computations will significantly enhance performance, while keeping the quality of results in a user-constrained range. This chapter will explore how AxC can improve the performance and energy efficiency of hardware accelerators in DL applications during inference and training.

cs.AR↗

arXiv2022

Flash-Cosmos: In-Flash Bulk Bitwise Operations Using Inherent Computation Capability of NAND Flash Memory

Bulk bitwise operations, i.e., bitwise operations on large bit vectors, are prevalent in a wide range of important application domains, including databases, graph processing, genome analysis, cryptography, and hyper-dimensional computing. In conventional systems, the performance and energy efficiency of bulk bitwise operations are bottlenecked by data movement between the compute units and the memory hierarchy. In-flash processing (i.e., processing data inside NAND flash chips) has a high potential to accelerate bulk bitwise operations by fundamentally reducing data movement through the entire memory hierarchy. We identify two key limitations of the state-of-the-art in-flash processing technique for bulk bitwise operations; (i) it falls short of maximally exploiting the bit-level parallelism of bulk bitwise operations; (ii) it is unreliable because it does not consider the highly error-prone nature of NAND flash memory. We propose Flash-Cosmos (Flash Computation with One-Shot Multi-Operand Sensing), a new in-flash processing technique that significantly increases the performance and energy efficiency of bulk bitwise operations while providing high reliability. Flash-Cosmos introduces two key mechanisms that can be easily supported in modern NAND flash chips: (i) Multi-Wordline Sensing (MWS), which enables bulk bitwise operations on a large number of operands with a single sensing operation, and (ii) Enhanced SLC-mode Programming (ESP), which enables reliable computation inside NAND flash memory. We demonstrate the feasibility of performing bulk bitwise operations with high reliability in Flash-Cosmos by testing 160 real 3D NAND flash chips. Our evaluation shows that Flash-Cosmos improves average performance and energy efficiency by 3.5x/32x and 3.3x/95x, respectively, over the state-of-the-art in-flash/outside-storage processing techniques across three real-world applications.

cs.AR↗

arXiv2022

Hermes: Accelerating Long-Latency Load Requests via Perceptron-Based Off-Chip Load Prediction

Long-latency load requests continue to limit the performance of high-performance processors. To increase the latency tolerance of a processor, architects have primarily relied on two key techniques: sophisticated data prefetchers and large on-chip caches. In this work, we show that: 1) even a sophisticated state-of-the-art prefetcher can only predict half of the off-chip load requests on average across a wide range of workloads, and 2) due to the increasing size and complexity of on-chip caches, a large fraction of the latency of an off-chip load request is spent accessing the on-chip cache hierarchy. The goal of this work is to accelerate off-chip load requests by removing the on-chip cache access latency from their critical path. To this end, we propose a new technique called Hermes, whose key idea is to: 1) accurately predict which load requests might go off-chip, and 2) speculatively fetch the data required by the predicted off-chip loads directly from the main memory, while also concurrently accessing the cache hierarchy for such loads. To enable Hermes, we develop a new lightweight, perceptron-based off-chip load prediction technique that learns to identify off-chip load requests using multiple program features (e.g., sequence of program counters). For every load request, the predictor observes a set of program features to predict whether or not the load would go off-chip. If the load is predicted to go off-chip, Hermes issues a speculative request directly to the memory controller once the load's physical address is generated. If the prediction is correct, the load eventually misses the cache hierarchy and waits for the ongoing speculative request to finish, thus hiding the on-chip cache hierarchy access latency from the critical path of the off-chip load. Our evaluation shows that Hermes significantly improves performance of a state-of-the-art baseline. We open-source Hermes.

cs.AR↗

arXiv2021

Fast Exploration of Weight Sharing Opportunities for CNN Compression

The computational workload involved in Convolutional Neural Networks (CNNs) is typically out of reach for low-power embedded devices. There are a large number of approximation techniques to address this problem. These methods have hyper-parameters that need to be optimized for each CNNs using design space exploration (DSE). The goal of this work is to demonstrate that the DSE phase time can easily explode for state of the art CNN. We thus propose the use of an optimized exploration process to drastically reduce the exploration time without sacrificing the quality of the output.

cs.LG↗

arXiv2019

Exploration of Performance and Energy Trade-offs for Heterogeneous Multicore Architectures

Energy-efficiency has become a major challenge in modern computer systems. To address this challenge, candidate systems increasingly integrate heterogeneous cores in order to satisfy diverse computation requirements by selecting cores with suitable features. In particular, single-ISA heterogeneous multicore processors such as ARM big.LITTLE have become very attractive since they offer good opportunities in terms of performance and power consumption trade-off. While existing works already showed that this feature can improve system energy-efficiency, further gains are possible by generalizing the principle to higher levels of heterogeneity. The present paper aims to explore these gains by considering single-ISA heterogeneous multicore architectures including three different types of cores. For this purpose, we use the Samsung Exynos Octa 5422 chip as baseline architecture. Then, we model and evaluate Cortex A7, A9, and A15 cores using the gem5 simulation framework coupled to McPAT for power estimation. We demonstrate that varying the level of heterogeneity as well as the different core ratio can lead to up to 2.3x gains in energy efficiency and up to 1.5x in performance. This study further provides insights on the impact of workload nature on performance/energy trade-off and draws recommendations concerning suitable architecture configurations. This contributes in fine to guide future research towards dynamically reconfigurable HSAs in which some cores/clusters can be disabled momentarily so as to optimize certain metrics such as energy efficiency. This is of particular interest when dealing with quality-tunable algorithms in which accuracy can be then traded for compute effort, thereby enabling to use only those cores that provide the best energy-efficiency for the chosen algorithm.

cs.DC↗

arXiv2016

A Workflow for Fast Evaluation of Mapping Heuristics Targeting Cloud Infrastructures

Resource allocation is today an integral part of cloud infrastructures management to efficiently exploit resources. Cloud infrastructures centers generally use custom built heuristics to define the resource allocations. It is an immediate requirement for the management tools of these centers to have a fast yet reasonably accurate simulation and evaluation platform to define the resource allocation for cloud applications. This work proposes a framework allowing users to easily specify mappings for cloud applications described in the AMALTHEA format used in the context of the DreamCloud European project and to assess the quality for these mappings. The two quality metrics provided by the framework are execution time and energy consumption.

cs.DC↗