Memory allocation for embedded systems with a compile-time-unknown scratch-pad size

Abstract

This paper presents the first memory allocation scheme for embedded systems having a scratch-pad memory(SPM) whose size is unknown at compile-time. All existing memory allocation schemes for SPM require the SPM size to be known at compile-time; therefore tie the resulting executable to that size of SPM and not portable to other platforms having different SPM sizes. As size-portable code is valuable in systems supporting downloaded codes, our work presents a compiler method whose resulting executable is portable across SPMs of any size.Our technique is to employ a customized installer software, which decides the SPM allocation just before the program's first run, then modifies the program executable accordingly to implement the decided SPM allocation. Results show that our benchmarks average a 41% speedup versus an all-DRAM allocation, with overheads of 1.5% in code-size, 2% in run-time, and 3% in compile-time for our benchmarks. Meanwhile, an unrealistic upper-bound is approximated only slightly faster at 45% better than all-DRAM.

Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator

“…In this topic, a few research works about reconfigurable computing do introduce techniques for managing the memory and reducing the number of data transfers [22,3,29]. Still, the most active researches are about scratch-pad memory [18,16,27] or GPUs [2,6]. These approaches usually support single or shared memory organizations, and have various contributions like compile-time or operating-system-based allocation and copy policies [18,27,29], new memory allocators [16], or schedule-based optimizations for reducing the cost of data transfers [22,2,6].…”

“…Still, the most active researches are about scratch-pad memory [18,16,27] or GPUs [2,6]. These approaches usually support single or shared memory organizations, and have various contributions like compile-time or operating-system-based allocation and copy policies [18,27,29], new memory allocators [16], or schedule-based optimizations for reducing the cost of data transfers [22,2,6]. They are complementary to our approach since it supports run-time decisions and targets a distributed memory organization where the datapath tiles can only access their local memories.…”

Runtime Memory Allocation in a Heterogeneous Reconfigurable Platform

“…At the boundary between static and dynamic allocation, [6] performs a load time optimization that places the stack data in one of the memories using information computed at compile time, but taking into account the size of the memory at runtime. Compared to this approach, we manage both heap and stack data.…”

Scratchpad memory based power efficient optimization for MPSoC

“…During the execution, the different regions were scheduled to different cores. In further work in [19], the lengths of sub-loop divided were limited to balance the overloads of different cores for better performance of the whole system. The work in [20] proposed a complete share approach.…”