Application autotuning to support runtime adaptivity in multicore architectures

Abstract

In this work, we introduce an application autotuning framework to dynamically adapt applications in multicore architectures. In particular, the framework exploits design-time knowledge and multi-objective requirements expressed by the user, to drive the autotuning process at the runtime. It also exploits a monitoring infrastructure to get runtime feed-back and to adapt to external changing conditions. The intrusiveness of the autotuning framework in the application (in terms of refactoring and lines of code to be added) has been kept limited, also to minimize the integration cost. To assess the proposed framework, we carried out an experimental campaign to evaluate the overhead, the relevance of the described features and the efficiency of the framework

An Efficient Monte Carlo-Based Probabilistic Time-Dependent Routing Calculation Targeting a Server-Side Car Navigation System

“…On one hand we use the mARGOt [35] framework to dynamically tune the application, thus implementing the 1 // Load data 2 Routing::MCSimulation mc(edgesPath, profilePath); 3 auto run result = mc.RunMonteCarloSimulation(samples, startTime); 4 ResultStats stats(run result); 5 Routing::Data::WriteResultSingle(run result, outputFile); 6 return 0; Listing 1: Original source code before integrating the adaptivity layer adaptivity concepts presented in Section 4 and thus transforming the target application as highlighted in Figure 3.…”

The ANTAREX domain specific language for high performance computing

“…Other parameters (from line 4) are provided to improve several memoization approaches. For examples, the user can specify (1) the policy in case of conflicts regarding the same (lines 14-15), the method is searched (lines [17][18][19][20][21][22][23][24]. Then, in case of success, the code of the wrapper is added (line 28) to produce the memoization library, and (line 30) the code of the application is modified for calling the created "wrapper", this wrapper is also declared as a new method of the class.…”

“…, k n ) are software-knobs that modify the behavior of the application (e.g., parallelism level or the number of trials in a MonteCarlo simulation). The main goal of mARGOt 6 [23] is to enhance an application with an adaptive layer, aiming at tuning the software knobs to satisfy the application requirements at runtime. To achieve this goal, the mARGOt dynamic autotuning framework developed in ANTAREX is based on the MAPE-K feedback loop [24].…”

Supporting the Scale-Up of High Performance Application to Pre-Exascale Systems: The ANTAREX Approach

“…In ANTAREX, at runtime, the mARGOt tool [23] configures the available software knobs (application parameters, code transformations and code variants) according to the runtime information coming from application self-monitoring and system monitoring, thus creating an autotuning control loop. Finally, the runtime power manager, PowerCapper, is used to control the resource usage for the underlying computing infrastructure given the changing conditions [24,25].…”