Search
Results
-
Cascade: CPU Fuzzing via Intricate Program Generation
(2024)Generating interesting test cases for CPU fuzzing is akin to generating programs that exercise unusual states inside the CPU. The performance of CPU fuzzing is heavily influenced by the quality of these programs and by the overhead of bug detection. Our analysis of existing state-of-the-art CPU fuzzers shows that they generate programs that are either overly simple or execute a small fraction of their instructions due to invalid control ...Conference Paper -
-
GraphChef: Decision-Tree Recipes to Explain Graph Neural Networks
(2024)The Twelfth International Conference on Learning RepresentationsWe propose a new self-explainable Graph Neural Network (GNN) model: GraphChef. GraphChef integrates decision trees into the GNN message passing framework. Given a dataset, GraphChef returns a set of rules (a recipe) that explains each class in the dataset unlike existing GNNs and explanation methods that reason on individual graphs. Thanks to the decision trees, GraphChef recipes are human understandable. We also present a new pruning ...Conference Paper -
SoK: Efficient Design and Implementation of Polynomial Hash Functions over Prime Fields
(2024)2024 IEEE Symposium on Security and Privacy (SP)Poly1305 is a widely-deployed polynomial hash function. The rationale behind its design was laid out in a series of papers by Bernstein, the last of which dates back to 2005. As computer architectures evolved, some of its design features became less relevant, but implementers found new ways of exploiting these features to boost its performance. However, would we still converge to this same design if we started afresh with today’s computer ...Conference Paper -
Efficient and Scalable Graph Generation through Iterative Local Expansion
(2024)In the realm of generative models for graphs, extensive research has been conducted. However, most existing methods struggle with large graphs due to the complexity of representing the entire joint distribution across all node pairs and capturing both global and local graph structures simultaneously. To overcome these issues, we introduce a method that generates a graph by progressively expanding a single node to a target graph. In each ...Conference Paper -
AERO: Adaptive Erase Operation for Improving Lifetime and Performance of Modern NAND Flash-Based SSDs
(2024)International Conference on Architectural Support for Programming Languages and Operating Systems - ASPLOSThis work investigates a new erase scheme in NAND flash memory to improve the lifetime and performance of modern solid-state drives (SSDs). In NAND flash memory, an erase operation applies a high voltage (e.g., > 20 V) to flash cells for a long time (e.g., > 3.5 ms), which degrades cell endurance and potentially delays user I/O requests. While a large body of prior work has proposed various techniques to mitigate the negative impact of ...Conference Paper -
More Apps, Faster Hot-Launch on Mobile Devices via Fore/Background-aware GC-Swap Co-design
(2024)International Conference on Architectural Support for Programming Languages and Operating Systems - ASPLOSFaster app launching is crucial for the user experience on mobile devices. Apps launched from a background cached state, called hot-launching, have much better performance than apps launched from scratch. To increase the number of hot-launches, leading mobile vendors now cache more apps in the background by enabling swap. Recent work also proposed reducing the Java heap to increase the number of cached apps. However, this paper found that ...Conference Paper -
Skip It: Take Control of Your Cache!
(2024)International Conference on Architectural Support for Programming Languages and Operating Systems - ASPLOSMechanisms to explicitly manage the presence of data in caches are fundamental for the correctness and performance of modern systems. These operations, while critical, often incur significant performance penalties even when carefully used. Moreover, these mechanisms are implemented in proprietary and often undocumented hardware, so research into optimizations and novel designs is mostly limited to slow, simplified software simulations. ...Conference Paper -
Optimus: Warming Serverless ML Inference via Inter-Function Model Transformation
(2024)Conference Paper -
ML Training with Cloud GPU Shortages: Is Cross-Region the Answer?
(2024)EuroMLSys 2024 - Proceedings of the 2024 4th Workshop on Machine Learning and SystemsThe widespread adoption of ML has led to a high demand for GPU hardware and consequently, severe shortages of GPUs in the public cloud. Allocating a sufficient number of GPUs to train or fine-tune today’s large ML models in a single cloud region is often difficult. Users can get access to more GPUs if they are willing to run a ML training job using devices across different geographical regions. However, GPU nodes are connected with lower ...Conference Paper