Full-system analysis and characterization of interactive smartphone applications

Abstract

Smartphones have recently overtaken PCs as the primary consumer computing device in terms of annual unit shipments. Given this rapid market growth, it is important that mobile system designers and computer architects analyze the characteristics of the interactive applications users have come to expect on these platforms. With the introduction of highperformance, low-power, general purpose CPUs in the latest smartphone models, users now expect PC-like performance and a rich user experience, including high-definition audio and video, high-quality multimedia, dynamic web content, responsive user interfaces, and 3D graphics.In this paper, we characterize the microarchitectural behavior of representative smartphone applications on a currentgeneration mobile platform to identify trends that might impact future designs. To this end, we measure a suite of widely available mobile applications for audio, video, and interactive gaming. To complete this suite we developed BBench, a new fully-automated benchmark to assess a web-browser's performance when rendering some of the most popular and complex sites on the web. We contrast these applications' characteristics with those of the SPEC CPU2006 benchmark suite. We demonstrate that realworld interactive smartphone applications differ markedly from the SPEC suite. Specifically the instruction cache, instruction TLB, and branch predictor suffer from poor performance. We conjecture that this is due to the applications' reliance on numerous high level software abstractions (shared libraries and OS services). Similar trends have been observed for UI-intensive interactive applications on the desktop.

On-device objective-C application optimization framework for high-performance mobile processors

“…Unfortunately, the presence of dynamic dispatch can introduce significant overheads during method calls, which can directly impact execution time [2]. It has also been observed that smartphone applications suffer from increased code size and sparseness [3], and encounter higher branch mispredictions and instruction cache misses due to lack of instruction locality. As this paper will demonstrate, dynamic dispatch exacts a heavy toll due to the numerous control flow changes that occur from the large quantity of method call polymorphic sites.…”

EPE‐Mobile—A framework for early performance estimation of mobile applications

“…Figure 9 depicts an extract of the XML output file generated during the simulation of the synthetic program. In this example, one can observe the metrics obtained for the floating-point operation (lines [3][4][5][6][7][8][9][10][11][12][13][14]. For each operation, the parameters used in the reported simulations are informed (lines 4-7), and the collected metrics are presented (lines [8][9][10][11][12][13][14].…”

“…Apart from these benchmark applications, other well-known and open benchmark suites have also been used by platform providers and developers to compare different embedded platforms (processors and peripherals), [8][9][10] embedded processors alone, 11,12 or even programming language resources used in application development. 13,14 Since the source code of these benchmark suites is public, it is possible to know which performance metrics they exercise, thus allowing a more detailed analysis of the results obtained.…”

Panappticon: Event-based tracing to measure mobile application and platform performance

“…Gutierrez et al [16] developed a smartphone benchmark suite by characterizing representative smartphone applications. They found that smartphone applications have higher instruction cache miss rates than traditional SPEC benchmarks, and suggested increasing the cache size.…”