The Python programming language has established itself as a popular alternative for implementing scientific computing workflows. Its massive adoption across a wide spectrum of disciplines has created a strong community that develops tools for solving complex problems in science and engineering. In particular, there are several parallel programming libraries for Python codes that target multicore processors. We aim at comparing the performance and scalability of a subset of three popular libraries (Multiprocessing, PyMP, and Torcpy). We use the Particle-in-cell (PIC) method as a benchmark. This method is an attractive option for understanding physical phenomena, specially in plasma physics. A pre-existing PIC code implementation was modified to integrate Multiprocessing, PyMP, and Torcpy. The three tools were tested on a manycore and on a multicore processor by running different problem sizes. The results obtained consistently indicate that PyMP has the best performance, Multiprocessing showed a similar behavior but with longer execution times, and Torcpy did not properly scale when increasing the number of workers. Finally, a just-in-time (JIT) alternative was studied by using Numba, showing execution time reductions of up to 43%.
In the quest for extreme-scale supercomputers, the High Performance Computing (HPC) community has developed many resources (programming paradigms, architectures, method-ologies, numerical methods) to face the multiple challenges along the way. One of those resources are task-based parallel program-ming tools. The availability of mature programming models, pro-gramming languages, and runtime systems that use task-based parallelism represent a favorable ecosystem. The fundamental premise of these tools is their ability to naturally cope with dynamically changing execution conditions, i.e. adaptivity. In this paper, we explore Adaptive MPI, a parallel-object framework, as a mechanism to provide, among other features, automatic and dynamic load balancing for a particle-in-cell application. We ported a pre-existing MPI application on the Adaptive MPI infrastructure and highlight the changes required to the code. Our experimental results show Adaptive MPI has a minimum overhead, maintains the scalability of the original code, and it is able to alleviate an artificially-introduced load imbalance.
As the world moves away from traditional energy sources based on fossil fuels, several alternatives have been explored. One promising clean energy source is nuclear fusion. The fusion of hydrogen isotopes may provide generous consumable energy gains. However, nuclear fusion reactors are not ready to become a productive mechanism yet. To get a better understanding of plasma, numerical simulations and scientific visualizations over high-performance computing systems are mandatory. The results from the simulations and a proper display of the data are key to design and tune up nuclear fusion reactors. It is also thanks to the international collaboration effort such as the advisory contribution and tools of researchers from the Argonne National Laboratory in the United States in conjunction with the National Center for High Technology of Costa Rica that this work was successfully carried out. In a previous work, we explored a new approach of the scientific visualization of plasma confinement, presenting one model to generate realistic plasma representations. This work presents an evaluation of the expected quality of the images rendered with the created model. We propose a concept called visual plausibility as an evaluation attribute to rate each rendered image by physicists that already know about the plasma appearance.
The Western Central Valley (WCV) of Costa Rica is an area of interest, due to its high concentration of population and economic activity, presenting itself as a tropical monsoontype atmospheric basin (AB), with well-defined climatic seasons (dry and rainy). The present study proposes the assessment of low carbon steel (CS) atmospheric corrosion, based on ISO 9223 (2012) and associated standards. A general analysis of the atmospheric basin effect was initially performed on these data, followed by the basic modeling of air pollutants and meteorological parameters. The WCV is an area of low contamination, which corresponds to a C2 or C3 category, according to ISO 9223. It mainly shows significant climatic seasons (dry and rainy) effects on the initial corrosion rates, but obtaining similar annual corrosion results for them. The ISO 9223 annual atmospheric corrosion model overestimated the actual obtained corrosion values, whereas linear or logarithmic models gave better results, especially when time and/or time of wetness (TOW) were considered as variables.
Deep Learning (DL) applications are used to solve complex problems efficiently. These applications require complex neural network models composed of millions of parameters and huge amounts of data for proper training. This is only possible by parallelizing the necessary computations by so-called distributed deep learning (DDL) frameworks over many GPUs distributed over multiple nodes of a HPC cluster. These frameworks mostly utilize the compute power of the GPUs and use only a small portion of the available compute power of the CPUs in the nodes for I/O and inter-process communication, leaving many CPU cores idle and unused. The more powerful the base CPU in the cluster nodes, the more compute resources are wasted. In this paper, we investigate how much of this unutilized compute resources could be used for executing other applications without lowering the performance of the DDL frameworks. In our experiments, we executed a noise-generation application, which generates a very-high memory, network or I/O load, in parallel with DDL frameworks, and use HPC profiling and tracing techniques to determine whether and how the generated noise is affecting the performance of the DDL frameworks. Early results indicate that it might be possible to utilize the idle cores for jobs of other users without affecting the performance of the DDL applications in a negative way.
The profound impact of recent developments in artificial intelligence is unquestionable. The applications of deep learning models are everywhere, from advanced natural language processing to highly accurate prediction of extreme weather. Those models have been continuously increasing in complexity, becoming much more powerful than their original versions. In addition, data to train the models is becoming more available as technological infrastructures sense and collect more readings. Consequently, distributed deep learning training is often times necessary to handle intricate models and massive datasets. Running a distributed training strategy on a supercomputer exposes the models to all the considerations of a large-scale machine; reliability is one of them. As supercomputers integrate a colossal number of components, each fabricated on an ever decreasing feature-size, faults are common during execution of programs. A particular type of fault, silent data corruption, is troublesome because the system does not crash and does not immediately give an evident sign of an error. We set out to explore the effects of that type of faults by inspecting how distributed deep learning training strategies cope with bit-flips that affect their internal data structures. We used checkpoint alteration, a technique that permits the study of this phenomenon on different distributed training platforms and with different deep learning frameworks. We evaluated two distributed learning libraries (Distributed Data Parallel and Horovod) and found out Horovod is slightly more resilient to SDCs. However, fault propagation is similar in both cases, and the model is more sensitive to SDCs than the optimizer.
Accelerated computing is becoming more diverse as new vendors and architectures come into play. Although platform-specific programming models promise ease of development and better control over performance, they still restrict the portability of scientific applications. As the OpenMP offloading specification becomes adopted by more compilers, this programming model stands out as a vendor-neutral portable approach to heterogeneous programming. In this study, we port a plasma physics oriented field line tracing code from a CPU-based MPI+OpenMP approach to a GPU accelerated version, using OpenMP’s offloading capabilities. We analyze GPU performance across different vendors with respect to the original CPU version and test both prescriptive and descriptive approaches to accelerator programming. A maximum acceleration over the CPU implementation was achieved using OpenMP’s high-level offloading directives. In addition, we demonstrate portability across three different vendor GPUs with no code modifications.
Urothelial bladder cancer is the fourth most common malignancy in incidence in men, ranking eighth in mortality; while in women it is not in the top 10 in incidence or mortality. Advanced age is the main risk factor, with a median age at diagnosis between 70 and 84 years, with women 3 to 4 times less at risk than men. This is traditionally attributed to exposure and lifestyle, the use of carcinogens such as tobacco smoke, and exposure to carcinogenic aromatic amines is the second most important risk factor, representing (dyes, rubbers, textiles, paints and leathers) the second highest risk factor. Micro-invasive bladder cancer (MIBC) requires patients to undergo long-term invasive surveillance with diagnostic and therapeutic cystoscopy. The treatment of MIBC is undergoing rapid changes, as immunotherapy with checkpoint inhibitors, targeted therapies and antibody conjugates have become possible therapeutic options, and the identification of potential therapeutic targets with miRNAs is of increasing research interest.
Neotropical ecosystems are highly biodiverse; however, the excessive use of pesticides has polluted freshwaters, with deleterious effects on aquatic biota. This study aims to analyze concentrations of active ingredients (a.i) of pesticides and the risks posed to freshwater Neotropical ecosystems. We compiled information from 1036 superficial water samples taken in Costa Rica between 2009 and 2019. We calculated the detection frequency for 85 a.i. and compared the concentrations with international regulations. The most frequently detected pesticides were diuron, ametryn, pyrimethanil, flutolanil, diazinon, azoxystrobin, buprofezin, and epoxiconazole, with presence in >20% of the samples. We observed 32 pesticides with concentrations that exceeded international regulations, and the ecological risk to aquatic biota (assessed using the multi-substance potentially affected fraction model (msPAF)) revealed that 5% and 13% of the samples from Costa Rica pose a high or moderate acute risk, especially to primary producers and arthropods. Other Neotropical countries are experiencing the same trend with high loads of pesticides and consequent high risk to aquatic ecosystems. This information is highly valuable for authorities dealing with prospective and retrospective risk assessments for regulatory decisions in tropical countries. At the same time, this study highlights the need for systematic pesticide residue monitoring of fresh waters in the Neotropical region.
Turrialba is a stratovolcano located at the easternmost part of the Costa Rican volcanic front. After remaining quiescent for more than a century, in 1996 it started to show signs of unrest, until a first phreatomagmatic explosion occurred on January, 2010. Since then, the activity evolved from phreatic to magmatic, in a series of distinct eruptive phases. In this paper, we investigate the seismic records that span the whole eruptive process (2010-present), in order to identify precursory signals and characterize the volcanic evolution. A long-term analysis was carried out based on the continuous records, as well as seismic catalogs (volcano-tectonic seismicity, harmonic tremor, etc.). In addition, the gradual character of the evolution of this eruption allowed for the analysis of independent precursory stages. Thus, we inspected in detail the most important of those periods, particularly, prior to the first 2010 phreatomagmatic eruption, and prior to the 2016 transition to an open vent system. Temporary tremor amplitude decreases were found to precede most of the eruptive phases. In total, 9 pre-eruptive tremor abatement periods were identified spanning several days (5–44), which often concurred with a decrease in the SO2 flux. The analysis of the volcano-tectonic seismicity highlights the migration of magma from a deep (6–10 km) reservoir beneath the neighboring Irazú volcano towards Turrialba volcano, especially between the years 2015 and 2016. This activity peaked on December 2016 when a Mw 5.5 earthquake took place between both volcanoes. Harmonic tremor episodes thrived in the later phase when the system finally opened (2017–2018). In the short-term, compounded tonal seismic signals were identified as precursor events, such as long-period events followed by harmonic tremor or by a multichromatic coda similar to tornillo-type events. The co-occurrence of tremor amplitude decreases and tonal seismic signals is interpreted to be caused by a sealing of the hydroth