Search

As a new algorithm of graph embedding, graph neural networks (GNNs) have been widely used in many fields. However, GNN computing has the characteristics of both sparse graph processing and dense neural network, which make it difficult to be deployed efficiently on the existing graph processing accelerators or neural network accelerators. Recently, some GNN accelerators have been proposed, but the following challenges have not been fully solved: 1) the minibatch GNN inference scenario has the potential of software and hardware co-design, which can bring 30% computation amount reduction, and this is not well utilized. Besides, the cost of message flow graph construction is large and may account for more than 50% of the total delay; 2) the feature aggregation has a large amount of data access and relatively small amount of computation, which leads to low on-chip data reuse, only 10% of dense computing; and 3) without the optimization of sparse computing units, simple memory bank and cross bar architecture can easily lead to bank access conflict and load imbalance, reducing the utilization of computing units to less than 60%. In order to solve the above problems, we propose a algorithm-hardware co-design scheme to accelerate GNN inference, which includes three technologies: 1) a reuse-aware sampling method is proposed for minibatch inference scenarios, which reduces 30% of the calculation and improves the on-chip reusability of local data; 2) through the nodewise parallelism-aware quantization, the features and weights are quantized to integers with eight or four bits, which reduces the amount of memory access by at least four times; and 3) an accelerator supporting the above technologies is designed and evaluated, and different operations are supported by the sampling-inference integration architecture. The multibank on-chip memory pool is designed to support data reuse, and edge stream reordering is used to reduce data access conflicts, improving the utilization of computing units by $1.5\times $ . Combined with the above technologies, the experiments show that our design achieves $9.2\times $ speedup and $29\times $ energy efficiency improvement compared with the Deep Graph Library framework running on servers equipped with CPU and GPU.

The ever-increasing environmental risks posed by nanoplastics (NPs) have been drawing more concerns to their biological effects, but with limited information on pathogenic bacteria. This work reported the alterations in bacteriological properties of Escherichia coli O157:H7(E. coli O157:H7) under exposure to polystyrene (PS) NPs of different sizes. Results found that E. coli O157:H7growth was promoted by the addition of PS-NPs. At environmentally relevant exposure levels, E. coli O157:H7underwent metabolic adaptation after different PS-NP treatments, showing differentiated responsive volatile metabolic features. Moreover, enzymatic activity analysis suggested that E. coli O157:H7could address PS-NP-induced oxidative stress mainly viasupporting glutathione (GSH) synthesis. Furthermore, it could fuel the TCA cycle through highly expressing regulatory enzymes as well with a greater response to small PS-NPs. Also, there was an observed increase in the expression of the AcrAB-TolCefflux pump after the treatment of PS below 100 nm and with a possible consequence of heightening E. coli O157:H7resistance. Together, our findings provided the first look into the potential implications mediated by PS nanoparticles on E. coli O157:H7growth and development and in turn emphasized a potential increased contamination of E. coli O157:H7as a waterborne pathogen in PS-NP polluted water.

Deep Neural Network (DNN) INFerence-as-a-Service (INFaaS) is the dominating workload in current data centers, for which FPGAs become promising hardware platforms because of their high flexibility and energy efficiency. The dynamic and multi-tenancy nature of INFaaS requires careful design in three aspects: multi-tenant architecture, multi-DNN scheduling, and multi-core mapping. These three factors are critical to the system latency and energy efficiency but are also challenging to optimize since they are tightly coupled and correlated. This paper proposes H3M, an automatic Design Space Exploration (DSE) framework to jointly optimize the architecture, scheduling, and mapping for serving INFaaS on cloud FPGAs. H3M explores: (1) the architecture design space with Heterogeneous spatial Multi-tenant sub-accelerators, (2) layer-wise scheduling for Heterogeneous Multi-DNN workloads, and (3) single-layer mapping to the Homogeneous Multi-core architecture. H3M beats state-of-the-art multi-tenant DNN accelerators, Planaria and Herald, by up to 7.5× and 3.6× in Energy-Delay-Product (EDP) reduction on the ASIC platform. On the Xilinx U200 and U280 FPGA platforms, H3M offers 2.1-5.7× and 1.8-9.0× EDP reduction over Herald.

Convolutional neural networks (CNN) have become essential for many scientific and industrial applications, such as image classification and pattern detection. Among the devices that can implement neural networks, SRAM based FPGAs are a popular option due to their excellent parallel computing capability and good flexibility. However, SRAM-FPGAs are susceptible to radiation effects, which limits its application on safety critical applications. In this paper, the reliability of an accelerator based on the advanced Instruction-Set Architecture is evaluated based on hardware fault injection experiments. Each main module of the accelerator is evaluated separately, and the impact of parallelism and model features on the accelerator reliability is also examined. The experimental results reveal some important conclusions in terms of general hardware reliability and also of the particular model reliability. First, over 99% of SEUs on the computation modules will cause accuracy loss, and the reliability improves for higher parallelism. Second, a large portion of SEUs on the data mover and the instruction scheduler will cause system corruptions due to abnormal interactions with the ARM or other modules. Third, nonlinear activation and pooling layers are effective in reducing the effect of SEUs on computation modules, so models that use these layers tend to be more robust. The results provide a deep understanding of the impact of errors on CNNs implemented on ISA based FPGA accelerators (e.g., the Xilinx DPU).

We report the development of an automated genetic analyzer for human sample testing based on microfluidic rapid polymerase chain reaction (PCR) with high-resolution melting analysis (HRMA). The integrated DNA microfluidic cartridge was used on a platform designed with a robotic pipettor system that works by sequentially picking up different test solutions from a 384-well plate, mixing them in the tips, and delivering mixed fluids to the DNA cartridge. A novel image feedback flow control system based on a Canon 5D Mark II digital camera was developed for controlling fluid movement through a complex microfluidic branching network without the use of valves. The same camera was used for measuring the high-resolution melt curve of DNA amplicons that were generated in the microfluidic chip. Owing to fast heating and cooling as well as sensitive temperature measurement in the microfluidic channels, the time frame for PCR and HRMA was dramatically reduced from hours to minutes. Preliminary testing results demonstrated that rapid serial PCR and HRMA are possible while still achieving high data quality that is suitable for human sample testing.

The electrophoretic motion of two particles in a uniform electric field is analyzed using bispherical coordinates. Thin electric double layers on the particle surfaces and Stokes flow of the surrounding fluid are assumed. The particles may be of different sizes and zeta potentials, suspended in an unbounded quiescent Newtownian fluid. The particles are separated by an arbitrary distance, and their line of centers has arbitrary orientation with respect to the electric field. Extensive mathematical treatment is performed on the solutions of the electrostatic and hydrodynamic governing equations, which are solved using scalar stream functions, potential functions, and auxiliary functions in series forms. As a result, fast and accurate solutions are possible for very wide ranges of particle separations and size ratios, thereby extending previous work on electrophoretic mobility functions for interacting particles. The new code can be incorporated directly into other codes which require electrophoretic mobility functions. As an example, pairwise electrophoretic heteroaggregation rates are calculated using the new code.

The stability and pairwise aggregation rates of small spherical particles in a heterogeneous suspension under the collective effects of gravitational motion and electrophoretic migration are analyzed. The particles are assumed to be non-Brownian, with thin, unpolarized double layers and different zeta potentials. The gravity vector and the electric field are assumed to be oriented in either the same direction or opposite directions. The particle aggregation rates are always enhanced by the presence of an electric field for parallel alignment of the gravitational and electrophoretic velocities. For antiparallel alignment with the magnitude of the gravitational relative velocity exceeding the magnitude of the electrophoretic relative velocity between two widely separated particles, the particle aggregation rates are reduced by the presence of the electric field, and there is a “collision-forbidden” region in parameter space due to stronger hydrodynamic interactions of the particles for gravitational motion than for electrophoretic motion. For antiparallel alignment with the magnitude of the electrophoretic relative velocity exceeding the magnitude of the gravitational relative velocity between two widely separated particles, the particle aggregation rates are enhanced by the presence of the electric field.

Electrophoretic migration of sulfated polystyrene latex particle bands was investigated in a density-gradient electrophoresis apparatus. The bands spread with time, and their velocity was found to depend on particle concentration and to decrease during the initial stages of electrophoresis. A change in pH during electrophoresis occurred, and this change may have caused variations in band velocity.