Zheng Wang

Zheng Wang

Auto-generated profile, based on publicly available information

Zheng Wang

University of California San Diego

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 June 2023
Verified
Conference Paper
Open access
SLSiqi Li
Siqi Li
FTFengbin Tu
Fengbin Tu
LLLiu Liu
Liu Liu
JLJilan Lin
Jilan Lin
ZWZheng Wang
Zheng Wang
YXYuan Xie
Yuan Xie
With the rapid growth of classification scale in deep learning systems, the final classification layer becomes extreme classification with a memory footprint exceeding the main memory capacity of the CPU or GPU. The emerging in-storage-computing technique offers an opportunity on account of the fact that SSD has enough storage capacity for the parameters of extreme classification. However, the limited performance of naive in-storage-computing schemes is insufficient to support the heavy workload of extreme classification. To this end, we propose ECSSD, the first hardware/data layout co-designed in-storage-computing architecture for extreme classification, based on the approximate screening algorithm. We propose an alignment-free floating-point MAC circuit technique to improve the computational ability under the limited area budget of in-storage-computing schemes so that the computational ability can match SSD's high internal bandwidth. We present a heterogeneous data layout design for the 4/32-bit weight data in the approximate screening algorithm to avoid data transfer interference and further utilize the internal DRAM bandwidth of SSD. Moreover, we propose a learning-based adaptive interleaving framework to balance the access workload in each flash channel and improve channel-level bandwidth utilization. Putting them together, our ECSSD achieves 3.24--49.87x performance improvements compared with state-of-the-art baselines.

17 April 2024
Verified
Conference Paper
Open access
BFBoyuan Feng
Boyuan Feng
ZWZheng Wang
Zheng Wang
YWYuke Wang
Yuke Wang
SYShu Yang
Shu Yang
YDYufei Ding
Yufei Ding
Zero knowledge Neural Networks draw increasing attention for guaranteeing computation integrity and privacy of neural networks (NNs) based on zero-knowledge Succinct Non-interactive ARgument of Knowledge (zkSNARK) security scheme. However, the performance of zkSNARK NNs is far from optimal due to the million-scale circuit computation with heavy scalar-level dependency. In this paper, we propose a type-based optimizing framework for efficient zero-knowledge NN inference, namely ZENO (ZEro knowledge Neural network Optimizer). We first introduce ZENO language construct to maintain high-level semantics and the type information (e.g., privacy and tensor) for allowing more aggressive optimizations. We then propose privacy-type driven and tensor-type driven optimizations to further optimize the generated zkSNARK circuit. Finally, we design a set of NN-centric system optimizations to further accelerate zkSNARK NNs. Experimental results show that ZENO achieves up to 8.5× end-to-end speedup than state-of-the-art zkSNARK NNs. We reduce proof time for VGG16 from 6 minutes to 48 seconds, which makes zkSNARK NNs practical.

27 April 2024
Verified
Conference Paper
Open access
ZWZheng Wang
Zheng Wang
YWYuke Wang
Yuke Wang
JDJiaqi Deng
Jiaqi Deng
DZDa Zheng
Da Zheng
ALAng Li
Ang Li
YDYufei Ding
Yufei Ding
Ensuring high-quality recommendations for newly onboarded users requires the continuous retraining of Deep Learning Recommendation Models (DLRMs) with freshly generated data. To serve the online DLRM retraining, existing solutions use hundreds of CPU computing nodes designated for input preprocessing, causing significant power consumption that surpasses even the power usage of GPU trainers. To this end, we propose RAP, an end-to-end DLRM training framework that supports Resource-aware Automated GPU sharing for DLRM input Preprocessing and Training. The core idea of RAP is to accurately capture the remaining GPU computing resources during DLRM training for input preprocessing, achieving superior training efficiency without requiring additional resources. Specifically, RAP utilizes a co-running cost model to efficiently assess the costs of various input preprocessing operations, and it implements a resource-aware horizontal fusion technique that adaptively merges smaller kernels according to GPU availability, circumventing any interference with DLRM training. In addition, RAP leverages a heuristic searching algorithm that jointly optimizes both the input preprocessing graph mapping and the co-running schedule to maximize the end-to-end DLRM training throughput. The comprehensive evaluation shows that RAP achieves 1.99× speedup on average over the sequential GPU-based DLRM input preprocessing baseline. In addition, the end-to-end training throughput of RAP is only 3.24% lower than the ideal case, which has no input preprocessing overhead.

26 October 2026 in ACM Transactions on Architecture and Code Optimization
Verified
Journal Article
Open access
ZWZheng Wang
Zheng Wang
ZYZhongkai Yu
Zhongkai Yu
KWKaijian Wang
Kaijian Wang
YLYichen Lin
Yichen Lin
YLYikai Li
Yikai Li
YDYufei Ding
Yufei Ding
The embedding tables in Deep Learning Recommendation Models (DLRMs) require significant memory capacity and bandwidth but relatively lower computing power, making it economically inefficient to scale by adding more GPUs solely to meet memory requirements. Recent advances in Compute Express Link (CXL) and near-data processing (NDP) offer a promising path for expanding system memory to support DLRM training with large embedding tables. However, existing CXL+NDP designs are limited to single-GPU scenarios and fail to address key challenges in multi-GPU settings, like memory contention and embedding placement. To address this issue, we propose RecDM , an efficient training system for large-scale Rec ommendation models on D isaggregated M emory. RecDM features a modular, many-to-many CXL-based architecture with lightweight NDP units integrated into the CXL controller of each memory expansion unit, avoiding changes to DRAM chips or DIMM organization. To optimize training throughput, RecDM introduces a hierarchical memory device allocation strategy that balances memory bandwidth and capacity by combining shared and exclusive device mappings. Furthermore, RecDM proposes a bandwidth-driven 2D embedding table sharding method, enabling fine-grained placement across heterogeneous memory hierarchies and devices. Finally, RecDM incorporates an input-adaptive communication routing mechanism combined with a pipelined GPU-CXL execution model to reduce synchronization and data movement overhead. Comprehensive experiments demonstrate that RecDM achieves an average 10.2× speedup over prior work that offloads embedding tables to host DRAM, and outperforms state-of-the-art CXL+NDP solution by an average of 2.2×.