Jeng-Hau Lin

Hao Zhuang, Shih-Hung Weng, Jeng-Hau Lin et al.

We proposed MATEX, a distributed framework for transient simulation of power distribution networks (PDNs). MATEX utilizes matrix exponential kernel with Krylov subspace approximations to solve differential equations of linear circuit. First, the whole simulation task is divided into subtasks based on decompositions of current sources, in order to reduce the computational overheads. Then these subtasks are distributed to different computing nodes and processed in parallel. Within each node, after the matrix factorization at the beginning of simulation, the adaptive time stepping solver is performed without extra matrix re-factorizations. MATEX overcomes the stiff-ness hinder of previous matrix exponential-based circuit simulator by rational Krylov subspace method, which leads to larger step sizes with smaller dimensions of Krylov subspace bases and highly accelerates the whole computation. MATEX outperforms both traditional fixed and adaptive time stepping methods, e.g., achieving around 13X over the trapezoidal framework with fixed time step for the IBM power grid benchmarks.

Hao Zhuang, Wenjian Yu, Shih-Hung Weng et al.

We design an algorithmic framework using matrix exponentials for time-domain simulation of power delivery network (PDN). Our framework can reuse factorized matrices to simulate the large-scale linear PDN system with variable stepsizes. In contrast, current conventional PDN simulation solvers have to use fixed step-size approach in order to reuse factorized matrices generated by the expensive matrix decomposition. Based on the proposed exponential integration framework, we design a PDN solver R-MATEX with the flexible time-stepping capability. The key operation of matrix exponential and vector product (MEVP) is computed by the rational Krylov subspace method. To further improve the runtime, we also propose a distributed computing framework DR-MATEX. DR-MATEX reduces Krylov subspace generations caused by frequent breakpoints from a large number of current sources during simulation. By virtue of the superposition property of linear system and scaling invariance property of Krylov subspace, DR-MATEX can divide the whole simulation task into subtasks based on the alignments of breakpoints among those sources. The subtasks are processed in parallel at different computing nodes without any communication during the computation of transient simulation. The final result is obtained by summing up the partial results among all the computing nodes after they finish the assigned subtasks. Therefore, our computation model belongs to the category known as Embarrassingly Parallel model. Experimental results show R-MATEX and DR-MATEX can achieve up to around 14.4X and 98.0X runtime speedups over traditional trapezoidal integration based solver with fixed timestep approach.

Jeng-Hau Lin, Yunfan Yang, Rajesh Gupta et al.

Memory and computation efficient deep learning architec- tures are crucial to continued proliferation of machine learning capabili- ties to new platforms and systems. Binarization of operations in convo- lutional neural networks has shown promising results in reducing model size and computing efficiency. In this paper, we tackle the problem us- ing a strategy different from the existing literature by proposing local binary pattern networks or LBPNet, that is able to learn and perform binary operations in an end-to-end fashion. LBPNet1 uses local binary comparisons and random projection in place of conventional convolu- tion (or approximation of convolution) operations. These operations can be implemented efficiently on different platforms including direct hard- ware implementation. We applied LBPNet and its variants on standard benchmarks. The results are promising across benchmarks while provid- ing an important means to improve memory and speed efficiency that is particularly suited for small footprint devices and hardware accelerators.

Jeng-Hau Lin, Tianwei Xing, Ritchie Zhao et al.

State-of-the-art convolutional neural networks are enormously costly in both compute and memory, demanding massively parallel GPUs for execution. Such networks strain the computational capabilities and energy available to embedded and mobile processing platforms, restricting their use in many important applications. In this paper, we push the boundaries of hardware-effective CNN design by proposing BCNN with Separable Filters (BCNNw/SF), which applies Singular Value Decomposition (SVD) on BCNN kernels to further reduce computational and storage complexity. To enable its implementation, we provide a closed form of the gradient over SVD to calculate the exact gradient with respect to every binarized weight in backward propagation. We verify BCNNw/SF on the MNIST, CIFAR-10, and SVHN datasets, and implement an accelerator for CIFAR-10 on FPGA hardware. Our BCNNw/SF accelerator realizes memory savings of 17% and execution time reduction of 31.3% compared to BCNN with only minor accuracy sacrifices.