Open access to the scientific literature means the removal of barriers (including price barriers) from accessing scholarly work. There are two parallel ┬ôroads┬ö towards open access: Open Access articles and self-archiving. Open Access articles are immediately, freely available on their Web site, a model mostly funded by charges paid by the author (usually through a research grant). The alternative for a researcher is ┬ôself-archiving┬ö (i.e., to publish in a traditional journal, where only subscribers have immediate access, but to make the article available on their personal and/or institutional Web sites (including so-called repositories or archives)), which is a practice allowed by many scholarly journals.
open access raises practical and policy questions for scholars, publishers, funders, and policymakers alike, including what the return on investment is when paying an article processing fee to publish in an Open Access articles, or whether investments into institutional repositories should be made and whether self-archiving should be made mandatory, as contemplated by some funders.
Exascale systems, featuring the capability of executing quintillion 1018 operations per second, are expected to be deployed in 2018 andwill bring significant advancements in a number of scientific fields of immediate global importance (such as medicine, biology, national security, and energy). Exascale platforms will be qualitatively different from current high-performance computing systems. The main driving force for the growth in computational power will be the increase of parallelism on-chip. It is expected that over the next decade the number of nodes will grow by a factor of 10├â┬Ś while on-chip parallelism will grow by a factor of 100├â┬Ś. Adapting software applications for exascale computing will be difficult as the architectural complexity of exascale systems will be extremely high in terms of their degree of concurrency and heterogeneity, sensitivity to communications and data movement, and requirements for locality. The vision of this work is to allow for the support exascale application development by enabling advanced simulations of inter-node communication patterns and by engineering the tools for fast and effective intranode synchronization and resource sharing. To achieve this goal the we will: Design and implement automatic extraction of application skeletons for simulation analysis of internode communication using SST/macro. Performance analysis of an application can be done in several ways with varying degrees of fidelity. One of the most cost-effective ways is to do a coarse-grained study of large- scale parallel applications through the use of program skeletons. The concept of a ┬ťskeleton┬Ł represents an abstracted program that is derived from a larger program where source code that is determined to be irrelevant is removed. In this work, we extend our prior work on using compiler-based program analysis with ROSE to develop a semi-automatic approach for extracting program skeletons by employing the Program Dependance Graph (PDG) for our analysis. Introduce and apply a new methodology for large-scale simulation validation based on the executions statistical characteristics. Validation is highly important in parallel application simulations with a large number of parameters, a process that can vary depending on the structure of the simulator and the granularity of the models used. Common practice involves calculating the percentage error between the projected and the real execution time of a benchmark program. This coarse-grained approach often suffers from a parameter insensitivity problem in regions of high-dimensional parameter space. We will develop a validation tool set that aims to capture fine-grained execution details. It consists of a trace analysis tool that decomposes execution time into finer granularity, a trace comparison tool that quantizes the disparity between correspondent metrics of two executions, and a visualization tool that renders the analysis and comparison results in graphs. The analysis process will take into account five groups of statistical data profiled from program traces: overall traffic and timing, per-node traffic and timing, MPI function histogram, collective synchronization and node-to-node communication.
Last date updated on February, 2021