Chenfeng Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Chenfeng Zhao has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Roger D. Chamberlain
Department of Computer Science and Engineering, Washington University in St. Louis, St. Louis, MO 63130, USA
4 shared publications - Xuan Zhang
Northeastern University, Boston, MA, USA
4 shared publications - Arabella L. Garcia
Washington University School of Medicine in St. Louis
1 shared publication - Benjamin A. Garcia
Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, St. Louis, USA
1 shared publication - Clayton J. Faber
Computer Science and Engineering, Washington University in St. Louis, St. Louis, United States
1 shared publication - Emily Zahn
Washington University School of Medicine in St. Louis
1 shared publication
Latest Publications
- Open Access
- 17 December 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 18 (1) , 1-26
HLPerf: Demystifying the Performance of HLS-based Graph Neural Networks with Dataflow Architectures
The development of FPGA-based applications using HLS is fraught with performance pitfalls and large design space exploration times. These issues are exacerbated when the application is complicated and its performance is dependent on the input dataset, as is often the case with graph neural network approaches to machine learning. Here, we introduce HLPerf, an open-source, simulation-based performance evaluation framework for dataflow architectures that both supports early exploration of the design space and shortens the performance evaluation cycle. We apply the methodology to GNNHLS, an HLS-based graph neural network benchmark containing six commonly used graph neural network models and four datasets with distinct topologies and scales. The results show that HLPerf achieves over 10, 000× average simulation acceleration relative to RTL simulation and over 400× acceleration relative to state-of-the-art cycle-accurate tools at the cost of 7% mean error rate relative to actual FPGA implementation performance. This acceleration positions HLPerf as a viable component in the design cycle.
- 17 December 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 18 (1) , 1-26
Open AccessHLPerf: Demystifying the Performance of HLS-based Graph Neural Networks with Dataflow Architectures
The development of FPGA-based applications using HLS is fraught with performance pitfalls and large design space exploration times. These issues are exacerbated when the application is complicated and its performance is dependent on the input dataset, as is often the case with graph neural network approaches to machine learning. Here, we introduce HLPerf, an open-source, simulation-based performance evaluation framework for dataflow architectures that both supports early exploration of the design space and shortens the performance evaluation cycle. We apply the methodology to GNNHLS, an HLS-based graph neural network benchmark containing six commonly used graph neural network models and four datasets with distinct topologies and scales. The results show that HLPerf achieves over 10, 000× average simulation acceleration relative to RTL simulation and over 400× acceleration relative to state-of-the-art cycle-accurate tools at the cost of 7% mean error rate relative to actual FPGA implementation performance. This acceleration positions HLPerf as a viable component in the design cycle.
- Open Access
- 7 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 115-125
HLS Taking Flight: Toward Using High-Level Synthesis Techniques in a Space-Borne Instrument
FPGAs are widely deployed on high-energy astrophysics telescopes to preprocess and reduce sensor data read out by front-end electronics. Across instruments, these computational pipelines have similar semantics, sharing common stages such as pedestal subtraction, signal integration, zero-suppression, island detection, and centroiding. However, diverse telescope designs require unique implementations of these algorithms, and the logic is often rewritten from scratch for a new instrument. As an alternative, High-Level Synthesis (HLS) tools enable these algorithms to be implemented in a high-level language, which eases modifications and enables fast prototyping and deployment. Nonetheless, writing performant HLS code requires augmentation of the code with compiler-specific pragmas. In this work, we illustrate these challenges in the context of the Advanced Particle-astrophysics Telescope (APT), a proposed space-based observatory for gamma-ray sources, and its Antarctic Demonstrator (ADAPT). We implement its front-end algorithms using HLS, demonstrate the use of pragmas to enable optimizations, then explore speed and area tradeoffs, which are especially important given the limited power budget afforded by a satellite instrument. We demonstrate that with HLS, ADAPT will be able to process scintillating tile data from 200 000 gamma-ray events per second.
- 7 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 115-125
Open AccessHLS Taking Flight: Toward Using High-Level Synthesis Techniques in a Space-Borne Instrument
FPGAs are widely deployed on high-energy astrophysics telescopes to preprocess and reduce sensor data read out by front-end electronics. Across instruments, these computational pipelines have similar semantics, sharing common stages such as pedestal subtraction, signal integration, zero-suppression, island detection, and centroiding. However, diverse telescope designs require unique implementations of these algorithms, and the logic is often rewritten from scratch for a new instrument. As an alternative, High-Level Synthesis (HLS) tools enable these algorithms to be implemented in a high-level language, which eases modifications and enables fast prototyping and deployment. Nonetheless, writing performant HLS code requires augmentation of the code with compiler-specific pragmas. In this work, we illustrate these challenges in the context of the Advanced Particle-astrophysics Telescope (APT), a proposed space-based observatory for gamma-ray sources, and its Antarctic Demonstrator (ADAPT). We implement its front-end algorithms using HLS, demonstrate the use of pragmas to enable optimizations, then explore speed and area tradeoffs, which are especially important given the limited power budget afforded by a satellite instrument. We demonstrate that with HLS, ADAPT will be able to process scintillating tile data from 200 000 gamma-ray events per second.
- Open Access
GNNHLS: Evaluating Graph Neural Network Inference via High-Level Synthesis
With the ever-growing popularity of Graph Neural Networks (GNNs), efficient GNN inference is gaining tremendous attention. Field-Programming Gate Arrays (FPGAs) are a promising execution platform due to their fine-grained parallelism, low-power consumption, reconfigurability, and concurrent execution. Even better, High-Level Synthesis (HLS) tools bridge the gap between the non-trivial FPGA development efforts and rapid emergence of new GNN models. In this paper, we propose GNNHLS, an open-source framework to comprehensively evaluate GNN inference acceleration on FPGAs via HLS, containing a software stack for data generation and baseline deployment, and FPGA implementations of 6 well-tuned GNN HLS kernels. We evaluate GNNHLS on 4 graph datasets with distinct topologies and scales. The results show that GNNHLS achieves up to 50.8x speedup and 423x energy reduction relative to the CPU baselines. Compared with the GPU baselines, GNNHLS achieves up to 5.16x speedup and 74.5x energy reduction.
Open AccessGNNHLS: Evaluating Graph Neural Network Inference via High-Level Synthesis
With the ever-growing popularity of Graph Neural Networks (GNNs), efficient GNN inference is gaining tremendous attention. Field-Programming Gate Arrays (FPGAs) are a promising execution platform due to their fine-grained parallelism, low-power consumption, reconfigurability, and concurrent execution. Even better, High-Level Synthesis (HLS) tools bridge the gap between the non-trivial FPGA development efforts and rapid emergence of new GNN models. In this paper, we propose GNNHLS, an open-source framework to comprehensively evaluate GNN inference acceleration on FPGAs via HLS, containing a software stack for data generation and baseline deployment, and FPGA implementations of 6 well-tuned GNN HLS kernels. We evaluate GNNHLS on 4 graph datasets with distinct topologies and scales. The results show that GNNHLS achieves up to 50.8x speedup and 423x energy reduction relative to the CPU baselines. Compared with the GPU baselines, GNNHLS achieves up to 5.16x speedup and 74.5x energy reduction.
- Open Access
- 5 September 2023
- Published by American Chemical Society (ACS) in ChemRxiv
SWAMNA: a comprehensive platform for analysis of nucleic acid modifications
The interest in MS-based analysis of modified nucleic acids is increasing due to the application of nucleic acids in therapeutics. However, there are few available integrated platforms for characterizing nucleic acid modifications. Herein, we report a general mass spectrometry-based SWATH platform to identify and quantify both RNA and DNA modifications, which we call SWATH Analysis of Modified Nucleic Acids (SWAMNA). SWAMNA incorporates the search engine, NuMo Finder, enabling the analysis of modifications in native and permethylated form. SWAMNA will aid discoveries that provide new insights into nucleic acid modifications.
- 5 September 2023
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessSWAMNA: a comprehensive platform for analysis of nucleic acid modifications
The interest in MS-based analysis of modified nucleic acids is increasing due to the application of nucleic acids in therapeutics. However, there are few available integrated platforms for characterizing nucleic acid modifications. Herein, we report a general mass spectrometry-based SWATH platform to identify and quantify both RNA and DNA modifications, which we call SWATH Analysis of Modified Nucleic Acids (SWAMNA). SWAMNA incorporates the search engine, NuMo Finder, enabling the analysis of modifications in native and permethylated form. SWAMNA will aid discoveries that provide new insights into nucleic acid modifications.
- Open Access
- 9 May 2023
- Published by Association for Computing Machinery (ACM)
- p. 42-51
SuperCut
The parallel execution of many graph algorithms is frequently dominated by data communication overheads between compute nodes. This bottleneck becomes even more pronounced in Near-Memory Processing (NMP) architectures with multiple memory cubes as local memory accesses are less expensive. Existing near-memory architectures typically use graph partitioning methods with a fixed vertex assignment, which limits their potential to improve performance and reduce energy consumption. Here, we argue that an NMP-based graph processing system should also consider the distribution of vertices onto memory cubes. We propose SuperCut, a framework for near-memory architectures to effectively reduce communication overheads while maintaining computational balance. We evaluate SuperCut via architectural simulation with 6 real-world datasets and 4 representative applications. The results show that it provides up to 1.8x total energy reduction and 2.6x speedup relative to current state-of-the-art approaches.
- 9 May 2023
- Published by Association for Computing Machinery (ACM)
- p. 42-51
Open AccessSuperCut
The parallel execution of many graph algorithms is frequently dominated by data communication overheads between compute nodes. This bottleneck becomes even more pronounced in Near-Memory Processing (NMP) architectures with multiple memory cubes as local memory accesses are less expensive. Existing near-memory architectures typically use graph partitioning methods with a fixed vertex assignment, which limits their potential to improve performance and reduce energy consumption. Here, we argue that an NMP-based graph processing system should also consider the distribution of vertices onto memory cubes. We propose SuperCut, a framework for near-memory architectures to effectively reduce communication overheads while maintaining computational balance. We evaluate SuperCut via architectural simulation with 6 real-world datasets and 4 representative applications. The results show that it provides up to 1.8x total energy reduction and 2.6x speedup relative to current state-of-the-art approaches.
- 31 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Design & Test
- Vol. 39 (2) , 65-73
Executing Data Integration Effectively and Efficiently Near the Memory
The article investigates the use of near-memory processing to perform data integration tasks. Based on the observation that most data integration workloads have regular memory access patterns and varying computation intensities, the authors develop a near-memory processing architecture to accelerate the execution of data transformation workloads and reduce their energy needs.
- 31 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Design & Test
- Vol. 39 (2) , 65-73
Executing Data Integration Effectively and Efficiently Near the Memory
The article investigates the use of near-memory processing to perform data integration tasks. Based on the observation that most data integration workloads have regular memory access patterns and varying computation intensities, the authors develop a near-memory processing architecture to accelerate the execution of data transformation workloads and reduce their energy needs.