A Quantitative Study of Memory System Interference in Chip Multiprocessor Architectures

Abstract

Abstract-The potential for destructive interference between running processes is increased as Chip Multiprocessors (CMPs) share more on-chip resources. We believe that understanding the nature of memory system interference is vital to achieve good fairness/complexity/performance trade-offs in CMPs. Our goal in this work is to quantify the latency penalties due to interference in all hardware-controlled, shared units (i.e. the onchip interconnect, shared cache and memory bus). To achieve this, we simulate a wide variety of realistic CMP architectures. In particular, we vary the number of cores, interconnect topology, shared cache size and off-chip memory bandwidth. We observe that interference in the off-chip memory bus accounts for between 63% and 87% of the total interference impact while the impact of cache capacity interference can be lower than indicated by previous studies (between 5% and 32% of the total impact). In addition, as much as 11% of the total impact can be due to uncontrolled allocation of shared cache Miss Status Holding Registers (MSHRs).

Dynamic management of TurboMode in modern multi-core chips

“…[16] quantifies destructive interference between separate processes on CMPs and concludes that most of the performance penalty comes from contention for the shared memory bus, supporting our strategy for detecting workload interference. [23] propose Cross-core interference Profiling Environment (CiPE) as a way of directly detecting cross-core interference in an accurate fashion, as opposed to indirect detection of autoturbo.…”

STHEM: Productive Implementation of High-Performance Embedded Image Processing Applications

“…Multi-threaded models. Current hardware platforms for image processing applications tend to contain multiple CPUs which means that the programmer may need to respond to the architectural challenges that can arise on multi-cores (e.g., [9,10,8]). Programming models for multi-cores has been studied extensively, and powerful tools such as OpenMP [6] are available to efficiently parallelise an application using task-based or data-parallel strategies.…”

TRP: A Foundational Platform for High-Performance Low-Power Embedded Image Processing

“…More specifically, we built the hardware platform around a Zynq UltraScale+ MPSoC which integrates a multi-core processor and an FPGA fabric on a single chip. Although adopting a multi-core platform is not without challenges (see e.g., [5,6,7]), these are outweighed by its performance and energy-efficiency advantages. Further, we only had the resources to develop a single hardware platform within the project which forced us to chose a platform that was acceptable for all target domains.…”