Modern mobile and embedded devices are required to be increasingly energy-efficient while running more sophisticated tasks, causing the CPU design to become more complex and employ more energy-saving techniques. This has created a greater need for fast and accurate power estimation frameworks for both run-time CPU energy management and design-space exploration. We present a statistically rigorous and novel methodology for building accurate run-time power models using performance monitoring counters (PMCs) for mobile and embedded devices, and demonstrate how our models make more efficient use of limited training data and better adapt to unseen scenarios by uniquely considering stability. Our robust model formulation reduces multicollinearity, allows separation of static and dynamic power, and allows a 100× reduction in experiment time while sacrificing only 0.6% accuracy. We present a statistically detailed evaluation of our model, highlighting and addressing the problem of heteroscedasticity in power modeling. We present software implementing our methodology and build power models for ARM Cortex-A7 and Cortex-A15 CPUs, with 3.8% and 2.8% average error, respectively. We model the behavior of the nonideal CPU voltage regulator under dynamic CPU activity to improve modeling accuracy by up to 5.5% in situations where the voltage cannot be measured. To address the lack of research utilizing PMC data from real mobile devices, we also present our data acquisition method and experimental platform software. We support this paper with online resources including software tools, documentation, raw data and further results.

As computer systems become increasingly complex, the need for fast and accurate simulation tools increases. Accurate but slow processor core models are often substituted with simple trace players to achieve faster memory-system simulation. However, existing trace-driven simulation techniques are limited in their applicability and availability. In this work, we capture elastic traces containing out-of-order core dependencies and effects of speculative execution, which overcome limitations of existing work. Additionally, we make our capture and replay modelling available in the gem5 simulator. Our trace-driven CPU achieves a speed-up of 6-8x compared to the reference core and predicts the performance with less than 1% error on average when the memory-system is changed.

Applications running on smartphones interact with the hardware and the system software differently, resulting in widely varying power consumption and hence thermal profiles. Typically, these smartphone platforms expose some hardware power control features to users, controlled through software governors such as cpufreq for dynamic voltage-frequency scaling (DVFS) and cpuquiet for dynamic core selection (DCS). Operating systems on these platforms manage these governors conservatively, independent of application's performance requirement. To address this, we propose an alternative approach, which uses reinforcement learning to explore the trade-off between power saving opportunities using DVFS and DCS and application's performance at run-time. The objective is to reduce power consumption, taking into consideration dynamic power, leakage power, and the inter-dependency between temperature and power. The reinforcement learning-based control is validated as a case-study on ARM A15-based nvidia's tegra smartphone through its implementation as a run-time manager (RTM). This RTM interfaces with different hardware performance counters and the embedded Linux Operating System through (1) the cpuquiet API to select cores at run-time; and (2) the cpufreq API to scale the frequency of active cores. Experiments with mobile and high performance applications demonstrate that the proposed approach achieves an average 22% (7-40%) power reduction compared to existing techniques.

As mobile devices gain ever more capabilities, their software and hardware complexity increases. Full system performance generally depends on complex hardware/software interactions, making it hard to reason about the performance impact of new components. These challenges are especially prominent for the flash storage as it involves a complex hardware architecture, and an extensive software stack. Universal Flash Storage (UFS) is an emerging flash interface proposed to address the growing demands of mobile workloads. However, due to the complexity of the system it is hard to determine the contribution of the storage device on the performance perceived by the user. To study the flash storage impact on modern mobile systems, and evaluate next-generation flash devices, we introduce a detailed UFS device model in the open-source full-system simulator gem5. We show the impact of different-performing flash devices on real mobile workloads, and compare the result with existing systems. Contrary to claims made by related work, we show that web browsing performance in itself is independent of the flash performance, but that tasks heavily utilizing the flash storage are clearly seeing the benefits of UFS. Our work enables performance analysis in both the hardware and the software layers of the storage system, and thus provides a platform for further research into mobile flash storage.

Compute requirements are increasing rapidly in systems ranging from mobile devices to servers. These, often massively parallel architectures, put increasing requirements on memory bandwidth and latency. The memory system greatly impacts both system performance and power, and it is key to capture the complex behaviour of the DRAM controller when evaluating CPU and GPU performance. By using full-system simulation, the interactions between the system components is captured. However, traditional DRAM controller models focus on modelling interactions between the controller and the DRAM rather than the interactions with the system. Moreover, the DRAM interactions are modelled on a cycle-by-cycle basis, leading to inflexibility and poor simulation performance. In this work, we present a high-level memory controller model, specifically designed for full-system exploration of future system architectures. Our event-based model is tailored to match a contemporary controller architecture, and captures the most important DRAM timing constraints for current and emerging DRAM interfaces, e.g. DDR3, LPDDR3 and WideIO. We show how our controller leverages the open-source gem5 simulation framework, and compare it to a state-of-the-art DRAM controller simulator. Our results show that our model is 7x faster on average, while maintaining the fidelity of the simulation. To highlight the capabilities of our model, we show that it can be used to evaluate a multi-processor memory system.

Dynamic Voltage and Frequency Scaling (DVFS) is an essential part of controlling the power consumption of any computer system, ranging from mobile phones to servers. DVFS efficiency relies on hardware-software co-optimization, thus using existing hardware cannot reveal the full optimization potential beyond the current implementation's characteristics. To explore the vast design space for DVFS efficiency, that straddles software and hardware, a simulation infrastructure must provide features that are not readily available today, for example: software controllable clock and voltage domains, support for the OS and the frequency scaling module of it, and an online power estimation methodology. As the main contribution, this work enables DVFS studies in a full-system simulator. We extend the gem5 simulator to support full-system DVFS modeling. By doing so, we enable energy-efficiency experiments to be performed in gem5 and we showcase such studies. Finally, we show that both existing and novel frequency governors for Linux and Android can be effortlessly integrated in the framework, and we evaluate the efficiency of different DVFS schemes.

The goals for the Æthereal network on silicon, as it was then called, were set in 2000 and its concepts were defined early 2001. Ten years on, what has been achieved? Did we meet the goals, and what is left of the concepts? In this paper we answer those questions, and evaluate different implementations, based on a new performance: cost analysis. We discuss and reflect on our experiences, and conclude with open issues and future directions.

To accommodate the growing number of applications integrated on a single chip, Networks on Chip (NoC) must offer scalability not only on the architectural, but also on the physical and functional level. In addition, real-time applications require Guaranteed Services (GS), with latency and throughput bounds. Traditionally, NoC architectures only deliver scalability on two of the aforementioned three levels, or do not offer GS. In this paper we present the composable and predictable aelite NoC architecture, that offers only GS, based on flit-synchronous Time Division Multiplexing (TDM). In contrast to other TDM-based NoCs, scalability on the physical level is achieved by using mesochronous or asynchronous links. Functional scalability is accomplished by completely isolating applications, and by having a router architecture that does not limit the number of service levels or connections. We demonstrate how aelite delivers the requested service to hundreds of simultaneous connections, and does so with 5 times less area compared to a state-of-the-art NoC.

Verification of application requirements is becoming a bottleneck in system-on-chip design, as the number of applications grows. Traditionally, the verification complexity increases exponentially with the number of applications and must be repeated if an application is added, removed, or modified. Predictable systems offering lower bounds on performance have been proposed to manage the increasing verification complexity, although this approach is only applicable to a restricted set of applications and systems. Composable systems, on the other hand, completely isolate applications in both the value and time domains, allowing them to be independently verified. However, existing approaches to composable system design are either restricted to applications that can be statically scheduled, or share resources using time-division multiplexing, which cannot efficiently satisfy tight latency requirements. In this paper, we present an approach to composable resource sharing based on latency-rate servers that supports any arbiter belonging to the class, providing a larger solution space for a given set of requirements. The approach can be combined with formal performance analysis using a variety of well-known modeling frame works. We furthermore propose an architecture for a resource front end that implements our concepts and provides composable service for any resource with bounded service time. The architecture supports both systems with buffers dimensioned to prevent overflow and systems with smaller buffers, where overflow is prevented with flow control. Finally, we experimentally demonstrate the usefulness of our approach with a simple use case sharing an SRAM memory.

A growing number of applications, often with firm or soft real-time requirements, are integrated on the same System on Chip, in the form of either hardware or software intellectual property. The applications are started and stopped at run time, creating different use-cases. Resources, such as interconnects and memories, are shared between different applications, both within and between use-cases, to reduce silicon cost and power consumption. The functional and temporal behaviour of the applications is verified by simulation and formal methods. Traditionally, designers resort to monolithic verification of the system as whole, since the applications interfere in shared resources, and thus affect each other's behaviour. Due to interference between applications, the integration and verification complexity grows exponentially in the number of applications, and the task to verify correct behaviour of concurrent applications is on the system designer rather than the application designers. In this work, we propose a Composable and Predictable Multi-Processor System on Chip (CoMPSoC) platform template. This scalable hardware and software template removes all interference between applications through resource reservations. We demonstrate how this enables a divide-and-conquer design strategy, where all applications, potentially using different programming models and communication paradigms, are developed and verified independently of one another. Performance is analyzed per application, using state-of-the-art dataflow techniques or simulation, depending on the requirements of the application. These results still apply when the applications are integrated onto the platform, thus separating system-level design and application design.

Networks on chip (NoC) have emerged as the design paradigm for scalable system on chip (SoC) communication infrastructure. Due to convergence, a growing number of applications are integrated on the same chip. When combined, these applications result in use-cases with different communication requirements. The NoC is configured per use-case and traditionally all running applications are disrupted during use-case transitions, even those continuing operation. In this paper we present a model that enables partial reconfiguration of NoCs and a mapping algorithm that uses the model to map multiple applications onto a NoC with undisrupted quality-of-service during reconfiguration. The performance of the methodology is verified by comparison with existing solutions for several SoC designs. We apply the algorithm to a mobile phone SoC with telecom, multimedia and gaming applications, reducing NoC area by more than 17% and power consumption by 50% compared to a state-of-the-art approach

Multi-processor systems on chip (MPSoC) platforms are becoming increasingly more heterogeneous and are shifting towards a more communication-centric methodology. Networks on chip (NoC) have emerged as the design paradigm for scalable on-chip communication architectures. As the system complexity grows, the problem emerges as how to design and instantiate such a NoC-based MPSoC platform in a systematic and automated way. This paper presents an integrated flow to automatically generate a highly configurable NoC-based MPSoC for FPGA instantiation. The system specification is done on a high level of abstraction, relieving the designer of error-prone and time consuming work. The flow uses the state-of-the-art /Ethereal NoC, and silicon hive processing cores, both configurable at design- and run-time. The authors use this flow to generate a range of sample designs whose functionality has been verified on a Celoxica RC300E development board. The board, equipped with a Xilinx Virtex II 6000, also offers a huge number of peripherals, and shows how the insertion is automated in the design for easy debugging and prototyping

Systems on chip (SoC) are becoming increasingly complex, with a large number of applications integrated on the same chip. Such a system often supports a large number of use-cases and is dynamically reconfigured when platform conditions or user requirements change. Networks on chip (NoC) offer the designer unsurpassed runtime flexibility. This flexibility stems from the programmability of the individual routers and network interfaces. When a change in use-case occurs, the application task graph and the network connections change. To mitigate the complexity in programming the many registers controlling the NoC, an abstraction in the form of a configuration library is needed. In addition, such a library must leave the modified system in a consistent state, from which normal operation can continue. In this paper we present the facilities for controlling change in a reconfigurable NoC. We show the architectural additions and the many trade-offs in the design of a run-time library for NoC reconfiguration. We qualitatively and quantitatively evaluate the performance, memory requirements, predictability and reusability of the different implementations

Networks-on-chip (NoC) are a scalable interconnect solution for systems on chip and are rapidly becoming reality. Monitoring is a key enabler for debugging or performance analysis and quality-of-service techniques. The NoC design problem and the NoC monitoring problem cannot be treated in isolation. We propose a monitoring-aware NoC design flow able to take into account the monitoring requirements in general. We illustrate our flow with a debug driven monitoring case study of transaction monitoring. By treating the NoC design and monitoring problems in synergy, the area cost of monitoring can be limited to 3-20% in general