Memory Allocation for Window-Based Image Processing on Multiple Memory Modules with Simple Addressing Functions

Memory-Access-Driven Context Partitioning for Window-Based Image Processing on Heterogeneous Multicore Processors

“…To solve this problem, addressing-function-constrained memory allocation is proposed in [5]. Since this paper is an extension of the work in [5], we briefly describe the addressing-function-constrained memory allocation.…”

“…Since this paper is an extension of the work in [5], we briefly describe the addressing-function-constrained memory allocation. The targeted application in [5] is window-based image processing. Window-based image processing refers to the processing of data in blocks.…”

“…When the transferred data amount decreases, the data transfer time also decreases. Memory allocation method based on the data sharing to reduce the data transfer time is already proposed in previous works [4], [5]. The work in [4] proposes a method to accelerate the optical-flow extraction by reducing the data transfer time.…”

“…The work in [4] proposes a method to accelerate the optical-flow extraction by reducing the data transfer time. The work in [5] proposes a memory allocation method for the image processing to reduce the data transfer time by increasing the data sharing. To get the maximum advantage from such works, we need an efficient context partitioning method to access the shared data.…”

“…The main contribution of this paper is to study of an efficient implementation of image processing in heterogeneous multicore processors. For this purpose, we extend the approach of the work in [5] to propose an efficient context partitioning method to access the shared data and to reuse the hardware resources. We discuss how to divide a large task in to partitions, how to map partitions to the reconfigurable accelerator by assigning them to contexts and how to schedule the context to effectively access the shared local memory data.…”