Optically reconfigurable processors
Optically reconfigurable processors
复制标题
光学可重构处理器
DOI：
10.1117/12.365910
复制
发表时间：
1999
期刊：
影响因子：
--
通讯作者：
D. Psaltis
中科院分区：
文献类型：
--
作者：
J. Mumbru;Gan Zhou;Suat U. Ay;X. An;G. Panotopoulos;F. Mok;D. Psaltis
可重构处理器，如现场可编程门阵列（FPGA），打开了新的计算范例，其中处理器能够定制其内部结构，以更好地实现给定的应用程序。典型的FPGA由可配置逻辑块阵列和完全可由用户编程以执行给定应用的互连网格组成。通过改变其内部连接，FPGA可以实现完全不同的新功能。然而，在大多数应用中，FPGA仅配置一次，并用作协处理器来执行一些高度复杂或耗时的计算。这种限制的原因是FPGA芯片和存储配置数据的外部存储器（通常是ROM）之间的通信带宽很小。光学可编程门阵列（OPGA）是传统FPGA的增强版，可以克服这个问题。OPGA利用由VCSEL阵列访问的全息存储器来编程其逻辑。片上逻辑已经补充了一个光电探测器阵列，以检测存储器中记录的配置模板。结合空间和移位多路复用来将配置页存储在存储器中，OPGA模块非常紧凑，并且具有极短的配置时间，允许动态重新配置。OPGA的重构能力可以应用于更有效地解决模式识别和数字分类问题。
Reconfigurable processors, like the field programmable gate arrays (FPGAs), open new computational paradigms where the processor is able to tailor its internal structure to better implement a given application. A typical FPGA consists of an array of configurable logic blocks and a mesh of interconnections fully programmable by the user to perform a given application. By just changing its internal connectivity, the FPGA can implement a totally different new function. However in most of the applications, the FPGA is configured only once and used as coprocessor to carry out some highly complex or time-consuming computation. The reason for such limitation is the small communication bandwidth between the FPGA chip and the external memory, usually ROM, where the configuration data is stored. The Optically Programmable Gate Array (OPGA), an enhanced version of a conventional FPGA, can overcome this problem. The OPGA utilizes a holographic memory accessed by an array of VCSELs to program its logic. The on-chip logic has been complemented with an array of photodetectors to detect the configuration template recorded in the memory. Combining spatial and shift multiplexing to store the configuration pages in the memory, the OPGA module is very compact and has extremely short configuration time allowing for dynamic reconfiguration. The reconfiguration capability of the OPGA can be applied to solve more efficiently problems in pattern recognition and digit classification.