Seongsik Park

State-based full predication for low power coarse-grained reconfigurable architecture

It has been one of the most fundamental challenges in architecture design to achieve high performance with low power while maintaining flexibility. Parallel architectures such as coarse-grained reconfigurable architecture, where multiple PEs are tightly coupled with each other, can be a viable solution to the problem. However, the PEs are typically controlled by a centralized control unit, which makes it hard to parallelize programs requiring different control of each PE. To overcome this limitation, it is essential to convert control flows into data flows by adopting the predicated execution technique, but it may incur additional power consumption. This paper reveals power issues in the predicated execution and proposes a novel technique to mitigate power overhead of predicated execution. Contrary to the conventional approach, the proposed mechanism can decide whether to suppress instruction execution or not without decoding the instructions and does not require additional instruction bits, thereby resulting in energy savings. Experimental results show that energy consumed by the reconfigurable array and its configuration memory is reduced by up to 23.9%.

An approach to code compression for CGRA

CGRA has been considered to be an attractive architecture for accelerating data-intensive applications due to the performance and flexibility that it can provide. However, the cache memory that stores the configuration code increases the silicon area significantly, making the architecture less attractive. This paper proposes an approach to saving the cache memory space through code compression for CGRA. It is based on the observation that typical configuration code consists of a repetition of same instruction patterns. Experiments with several applications show that the proposed approach reduces the code size by 56% on average and the required cache area by 26% on average and up to 68.6% when the hardware overhead is taken into account.

CloudSocket: Smart grid platform for datacenters

Todays datacenters are equipped with diverse computing and storage devices for handling a myriad of data and normally consume a significant amount of electrical energy. This paper proposes a smart grid inspired methodology to monitor and profile the energy consumption of a datacenter, with the aim of providing information useful for reducing the peak power consumption of the datacenter. Our energy measurement platform is named CloudSocket, and each CloudSocket unit can measure the power consumption of an individual computing node and periodically transmit the measurement information wirelessly to the coordinator unit that can manage many Cloud-Sockets simultaneously. We tested our methodology with a 32-node grid system that runs Apache Spark for large-scale data analytics. Analyzing our experimental results reveals how and where the peak power of each node in the grid overlaps, providing opportunities for informative coordination of the computing components for overall power reduction.

Techniques for improving coarse-grained reconfigurable architectures

This paper presents various novel techniques for improving coarse-grained reconfigurable architectures. Specifically, it presents techniques for supporting IEEE single precision floating-point standard, efficient handling of loop-carried dependency with variable-length FIFOs, efficient mapping of control flows, and sharing data with a host processor for transparent binary acceleration. Experiments with benchmark examples demonstrate the effectiveness of the proposed techniques.

CloudSocket: Fine-Grained Power Sensing and Analysis System for Datacenters

Today’s data centers have various computing and storage devices for processing a myriad of data, and they generally consume a considerable amount of electrical energy. This paper proposes a smart grid-inspired methodology to observe and profile the power consumption of a data center. Based on this technique, our paper provides information that is useful for moderating the peak power consumption of the data centers. Our power measurement platform consists of several devices named CloudSockets, and each CloudSocket unit can measure the power consumption of multiple computing nodes and periodically transmit measurement data wirelessly to the coordinator unit. This data can be used to analyze the relationship between the workload and the power consumption of the data center. We tested our methodology through the application of various algorithms with a 32-node distributed system that runs Apache Spark for large-scale data analytics. An analysis of our experimental results reveals how and where the peak power of each node in the grid overlaps, providing opportunities for informed coordination of the computing components for peak power reduction.

A Wall Induced Electro-Optic Effect at a Solid-Nematic Liquid Crystal Interface under an External Electric Field

Abstract A domain-wall is experimentally observed along a curved surface of an elongated glass spacer dispersed in a twisted nematic (TN) liquid crystal (LC) cell. The cylindrical optical fibers coated with two different polyimides (PI) for either homogeneous or homeotropic alignment are used as spacers in the TN cells. The formation of the walls/defects around the fiber spacers depends on the nature of the LC alignment at the fibers and the magnitude of the applied voltage. It is found that the domain-wall grows along the curved surface of the fiber spacer monotonically with increasing the electric field, and reaches a maximum around the Fredericks transition. The wall induced effect on the electro-optic performances of TN LC displays is also discussed.