Search for: All records

Memory latency and bandwidth are significant bottlenecks in designing in-memory indexes. Processing-in-memory (PIM), an emerging hardware design approach, alleviates this problem by embedding processors in memory modules, enabling low-latency memory access whose aggregated bandwidth scales linearly with the number of PIM modules. Despite recent work in balanced comparison-based indexes on PIM systems, building efficient tries for PIMs remains an open challenge due to tries' inherently unbalanced shape. This paper presents the PIM-trie, the first batch-parallel radix-based index for PIM systems that provides load balance and low communication under adversary-controlled workloads. We introduce trie matching-matching a query trie of a batch against the compressed data trie-as a key building block for PIM-friendly index operations. Our algorithm combines (i) hash-based comparisons for coarse-grained work distribution/elimination and (ii) bit-by-bit comparisons for fine-grained matching. Combined with other techniques (meta-block decomposition, selective recursive replication, differentiated verification), PIM-trie supports LongestCommonPrefix, Insert, and Delete in O(logP) communication rounds per batch and O(l/w) communication volume per string, where P is the number of PIM modules, l is the string length in bits, and w is the machine word size. Moreover, work and communication are load-balanced among modules whp, even under worst-case skew. 

Federated Learning (FL) under distributed concept drift is a largely unexplored area. Although concept drift is itself a well-studied phenomenon, it poses particular challenges for FL, because drifts arise staggered in time and space (across clients). Our work is the first to explicitly study data heterogeneity in both dimensions. We first demonstrate that prior solutions to drift adaptation, with their single global model, are ill-suited to staggered drifts, necessitating multiple-model solutions. We identify the problem of drift adaptation as a time-varying clustering problem, and we propose two new clustering algorithms for reacting to drifts based on local drift detection and hierarchical clustering. Empirical evaluation shows that our solutions achieve significantly higher accuracy than existing baselines, and are comparable to an idealized algorithm with oracle knowledge of the ground-truth clustering of clients to concepts at each time step. 

Batching has a fundamental influence on the efficiency of deep neural network (DNN) execution. However, for dynamic DNNs, efficient batching is particularly challenging as the dataflow graph varies per input instance. As a result, state-of-the-art frameworks use heuristics that result in suboptimal batching decisions. Further, batching puts strict restrictions on memory adjacency and can lead to high data movement costs. In this paper, we provide an approach for batching dynamic DNNs based on finite state machines, which enables the automatic discovery of batching policies specialized for each DNN via reinforcement learning. Moreover, we find that memory planning that is aware of the batching policy can save significant data movement overheads, which is automated by a PQ tree-based algorithm we introduce. Experimental results show that our framework speeds up state-of-the-art frameworks by on average 1.15x, 1.39x, and 2.45x for chain-based, tree-based, and lattice-based DNNs across CPU and GPU. The framework is open-sourced at https://github.com/gulang2019/ED-Batch.git. 

Federated Learning (FL) under distributed concept drift is a largely unexplored area. Although concept drift is itself a well-studied phenomenon, it poses particular challenges for FL, because drifts arise staggered in time and space (across clients). Our work is the first to explicitly study data heterogeneity in both dimensions. We first demonstrate that prior solutions to drift adaptation, with their single global model, are ill-suited to staggered drifts, necessitating multiple-model solutions. We identify the problem of drift adaptation as a time-varying clustering problem, and we propose two new clustering algorithms for reacting to drifts based on local drift detection and hierarchical clustering. Empirical evaluation shows that our solutions achieve significantly higher accuracy than existing baselines, and are comparable to an idealized algorithm with oracle knowledge of the ground-truth clustering of clients to concepts at each time step. 

The performance of today's in-memory indexes is bottlenecked by the memory latency/bandwidth wall. Processing-in-memory (PIM) is an emerging approach that potentially mitigates this bottleneck, by enabling low-latency memory access whose aggregate memory bandwidth scales with the number of PIM nodes. There is an inherent tension, however, between minimizing inter-node communication and achieving load balance in PIM systems, in the presence of workload skew. This paper presents PIM-tree , an ordered index for PIM systems that achieves both low communication and high load balance, regardless of the degree of skew in data and queries. Our skew-resistant index is based on a novel division of labor between the host CPU and PIM nodes, which leverages the strengths of each. We introduce push-pull search , which dynamically decides whether to push queries to a PIM-tree node or pull the node's keys back to the CPU based on workload skew. Combined with other PIM-friendly optimizations ( shadow subtrees and chunked skip lists ), our PIM-tree provides high-throughput, (guaranteed) low communication, and (guaranteed) high load balance, for batches of point queries, updates, and range scans. We implement PIM-tree, in addition to prior proposed PIM indexes, on the latest PIM system from UPMEM, with 32 CPU cores and 2048 PIM nodes. On workloads with 500 million keys and batches of 1 million queries, the throughput using PIM-trees is up to 69.7X and 59.1x higher than the two best prior PIM-based methods. As far as we know these are the first implementations of an ordered index on a real PIM system. 

The emergence of “robotics in the wild” has triggered a wave of recent research in hardware and software to boost robots’ compute capabilities. Nevertheless, research in this area is hindered by the lack of a comprehensive benchmark suite. In this paper, we present RTRBench, a benchmark suite for robotic kernels. RTRBench includes 16 kernels, spanning the entire software pipeline of a wide swath of robots, all implemented in C++ for fast execution. Together with the suite, we conduct an evaluation of the workloads at the architecture level. We pinpoint the sources of inefficiencies in a modern robotic processor when executing the robotic kernels, along with the opportunities for improvements. The source code of the benchmark suite is available in https://cmu-roboarch.github.io/rtrbench/. 

RACOD is an algorithm/hardware co-design for mobile robot path planning. It consists of two main components: CODAcc, a hardware accelerator for collision detection; and RASExp, an algorithm extension for runahead path exploration. CODAcc uses a novel MapReduce-style hardware computational model and massively parallelizes individual collision checks. RASExp predicts future path explorations and proactively computes its collision status ahead of time, thereby overlapping multiple collision detections. By affording multiple cheap CODAcc accelerators and overlapping collision detections using RASExp, RACOD significantly accelerates planning for mobile robots operating in arbitrary environments. Evaluations of popular benchmarks show up to 41.4× (self-driving cars) and 34.3× (pilotless drones) speedup with less than 0.3% area overhead. While the performance is maximized when CODAcc and RASExp are used together, they can also be used individually. To illustrate, we evaluate CODAcc alone in the context of a stationary robotic arm and show that it improves performance by 3.4×–3.8×. Also, we evaluate RASExp alone on commodity many-core CPU and GPU platforms by implementing it purely in software and show that with 32/128 CPU/GPU threads, it accelerates the end-to-end planning time by 8.6×/2.9×.