A post-compiler approach to scratchpad mapping of code

Abstract

ScratchPad Memories (SPMs) are commonly used in embedded systems because they are more energy-efficient than caches and enable tighter application control on the memory hierarchy. Optimally mapping code and data to SPMs is, however, still a challenge. This paper proposes an optimal scratchpad mapping approach for code segments, which has the distinctive characteristic of working directly on application binaries, thus requiring no access to either the compiler or the application source code -a clear advantage for legacy or proprietary, IP-protected applications.The mapping problem is solved by means of a Dynamic Programming algorithm applied to the execution traces of the target application. The algorithm is able to find the optimal set of instructions blocks to be moved into a dedicated SPM, either minimizing energy consumption or execution times. A patching tool, which can use the output of the optimal mapper, modifies the binary of the application and moves the relevant portions of its code segments to memory locations inside of the SPM.

Scratchpad memory management in a multitasking environment

“…Static SPM allocation techniques are presented in [1,3,19,27]. Except for Nguyen's work [19], all of these techniques require knowledge of the SPM size at compile time.…”

“…Except for Nguyen's work [19], all of these techniques require knowledge of the SPM size at compile time. Angiolini et al [1] present an SPM allocation scheme that selects code blocks which promise the highest energy savings using an algorithm based on Dynamic Programming. Banakar et al [3] solve the static assignment with a knapsack algorithm, both for code and data blocks.…”

“…Many studies have discussed code and/or data SPM allocation techniques [1,4,6,11,12,19,26,27]. Significant energy savings and performance improvements can be achieved by placing frequently accessed code/data blocks in the SPM.…”

“…However, most studies require the size of the SPM to be known at compile time. While a few consider multiple running tasks [1,26], none of the proposed techniques can handle dynamically created tasks.…”

WCET-aware data selection and allocation for scratchpad memory

“…Many SPM management approaches have been proposed. All the existing approaches can be classified into two categories: static allocation [1,6,10,11,22] and dynamic allocation [2, 3, 12-15, 20, 24, 27]. In the case of static allocation approaches, once an SPM resident is loaded into the SPM, its space in the SPM cannot be allocated to other SPM residents during the execution of the program.…”