Abstract:
PURPOSE: A reconfigurable processor and a method for handling interrupt thereof are provided to promptly process an interrupt request by securing some PE for interrupt handling when the interrupt request happens. CONSTITUTION: A CGA(Coarse-Grained Array)(101) includes plural PEs(Processing Elements), and a host processor(102) shares at least one PE with the CGA. A controller(103) designates at least one PE additionally. When an interrupt request happens while a loop operation is executed in the CGA, the controller allows the designated PE to process the interrupt request. A central register file(202) stores the processing result of the CGA and the host processor.
Abstract:
PURPOSE: An interrupt handling apparatus and method for an equal-model processor, and a processor including the interrupt handling apparatus are provided to promptly process an interrupt by processing an interrupt in the processor. CONSTITUTION: A remaining latency updater(14) compares a current latency with a residual latency. If the current latency is larger than the residual latency, the residual latency updating unit updates the residual latency to the value of the current latency. An interrupt support determiner(16) outputs a signal for indicating the interrupt support based on the residual latency. If the residual latency is larger than 1, the interrupt support determination unit outputs an interrupt non-support flag.
Abstract:
A profiler for optimizing architecture and application of a processor is provided to offer information needed for considering/optimizing a program and an application specific architecture processor in addition to the information for optimizing the program to a target processor. An architecture analyzer(231) generates architecture analysis information by analyzing architecture description(211) describing architecture of an application specific architecture processor including a plurality of PEs(Processing Elements). A static analyzer(232) generates static analysis information(222) by analyzing program static information describing static information of the program. A dynamic analyzer(233) generates dynamic analysis information(223) by analyzing program dynamic information describing dynamic information generated by simulating the program. A cross profiling analyzer(234) generates the information for optimizing the application specific architecture processor executing the program based on at least one of architecture, static, and dynamic analysis information. The architecture analyzer includes an operation analyzer generating operation group information for the PEs by analyzing the architecture description.
Abstract:
재구성 어레이의 동작 중 인터럽트 요청이 발생하면, 상기 재구성 어레이의 동작을 정지(pause)하는 단계, 상기 인터럽트 요청의 처리에 사용될 레지스터의 값을 저장하는 단계, 상기 인터럽트 요청에 대한 인터럽트 서비스를 수행하는 단계 및 상기 레지스터의 값을 복구하고, 상기 재구성 어레이의 동작을 재개(resume)하는 단계를 포함하는 재구성 어레이에서의 인터럽트 처리 방법이 제공된다. 코어스 그레인 어레이(coarse grained array), 재구성 아키텍처{reconfigurable architecture}, 인터럽트{interrupt}
Abstract:
A method and a device for performing multitasking in a reconfigurable array are provided to enable the reconfigurable array to quickly complete the multitasking for a plurality of reconfiguring operations by recovering only the unique information of a CPU in a time point when a main process is stopped and quickly performing the main process again. A unique memory(350) stores the unique information in response to a first control signal. A controller(340) verifies whether the reconfigurable operation is a main controlling process and generates a second control signal in response to a verification result. The reconfigurable array(310) receives a request for performing one reconfigurable operation during another reconfigurable operation, stops the reconfigurable operation and keeps the unique information of peripheral processors in response to the first control signal. The reconfigurable array receives the request for performing one reconfigurable operation during another reconfigurable operation, recovers the unique information of the CPU in response to the second control signal, and restarts the main processing operation.