The forge will collect the results from Fabrice Guet Ph.D. thesis (2015-2017), from Julien Durand stage (2018), and from Alfonso Mascarenas Gonzalez stage (2019) on the DIAGXTRM tool and the statistical analysis to timing analysis of single and multi-core embedded real-time systems.
The DIAGXTRM tool in its developed versions, trace of measurements to be analyzed with DIAGXTRM from the different collaborations (with Many-cores Kalray CISTER - Porto, GPUs CISTER - Porto, ISAE, industrial partners etc.), and the publications made during Fabrice guet thesis.
DIAGXTRM is also used for the statistical analysis under development.
The statistical analysis composes of first and second order statistical metrics and the Extreme Value Theory (all implemented into DIAGXTRM) for average and worst-case modeling of the system behavior. It bases on measurements of various system parameters and the statistics are applied for modeling.
The memory interference analysis is an application of the statistical analysis to identify, characterize, and validate interference into multi-core platoforms.
DIAGXTRM (from Fabrice Guet, Ph.D. thesis 2015-2017)
DIAGXTRM is a DIAGnostic tool for the eXTReMe value theory and is an innovative measurement-based probabilistic analysis conceived and developed by Fabrice Guet at the ONERA. It applies to both time-randomized and non time-randomized single-core, multi-core and many-core architectures for performance analyses of real-time embedded systems.
The forge will have an updated description of the tool with the tests applied and both the how to use the tool for trace probabilistic analysis and how to install it.
The need for the forge is to collect all the results achieved. The main ONERA users will manage all the info and publish such; an external user instead could download the tool for probabilistic timing analysis on trace of measurements. In particular, people from CISTER - Porto which are integrating DIAGXTRM into their analysis tool chain would need to download and test the latest versions.
The forge will be cited in the published paper so that reviewer or any user could test and validate the developed DIAGXTRM tool for probabilistic timing analysis.
diagXtrm tools are publically available, under a GPL licence.
Statistical analysis (works from Julien Durand stage in 2018)
Also, this forge contains the definition of statistical analysis for the study of interference.
The statistical analysis is for average and worst-case modeling of task execution time under different conditions. For average model it makes use of statistics such as mean, max, min, mode, standard deviation, etc.. For worst-case models, the analysis makes use of diagXtrm and the pWCET estimates.
The statistical analysis makes use of measurements and infers models of the system dynamics.
The statistical analysis has been applied to study interference on NXP T4240 and PikeOS hypervisor.
Memory interference (works from Alfonso Mascarenas Gonzalez stage in 2019)
To log into the project, where you will find a wiki containing install instructions, use :
- login: mauve
- password: !mauve