arXiv subjects

G. Taffoni

Publications and source records attributed to G. Taffoni.

15 recordsLinked to original sources

arXiv2025

Accelerating cosmological simulations on GPUs: a portable approach using OpenMP

In this work we present the porting to Graphics Processing Units (GPUs, using OpenMP target directives) and optimization of a key module within the cosmological {\pinocchio} code, a Lagrangian Perturbation Theory (LPT)-based framework widely used for generating dark matter (DM) halo catalogs. Our optimization focuses on a specific segment of the code responsible for calculating the collapse time of each particle involved in the simulation. Due to the embarrassingly parallel nature of this computation, it represents an ideal candidate for GPU offloading. As part of the porting process, we developed fully GPU-native implementations of both cubic spline and bilinear interpolation routines, required for evaluating collapse times. Since GNU Scientific Library (GSL) does not support GPU offloading, these custom implementations run entirely on the GPU and achieve residuals of only $\sim0.003\%$ when compared to the CPU-based implementation of GSL. Comparative benchmarking on the LEONARDO (NVIDIA-based) and SETONIX (AMD-based) supercomputers reveals notable portability and performance, with speedups of~\textit{4x} and up to~\textit{8x}, respectively. While collapse time calculation is not a primary bottleneck in the overall workflow, the acceleration reduces full production runs by $\sim 100$ seconds each leading to a cumulative saving of $\sim 160000$ Standard-h ($\sim28$ hours wall time) across thousands of simulations. Roofline analysis confirms that our GPU porting achieves over 80\% of the theoretical FP64 peak performance, confirming efficient compute-bound execution. This work demonstrates that OpenMP directives offer a portable, effective strategy for accelerating large-scale cosmological simulations on heterogeneous hardware.

astro-ph.IM↗

arXiv2023

A three-component giant radio halo: the puzzling case of the galaxy cluster Abell 2142

Turbulence introduced into the intra-cluster medium (ICM) through cluster merger events transfers energy to non-thermal components, and can trigger the formation of diffuse synchrotron radio sources. Typical diffuse sources in the forms of giant radio halos and mini-halos are found in merging and relaxed cool core galaxy clusters, respectively. On the other hand, recent observations have revealed an increasing complexity of the non-thermal phenomenology. Abell 2142 (A2142) is a mildly disturbed cluster that exhibits uncommon thermal and non-thermal properties. It is known to host a hybrid halo consisting of two components (H1 and H2), namely a mini-halo-like and an enigmatic elongated radio halo-like structure. We aim to investigate the properties, origin, and connections of each component. We present deep LOFAR observations of A2142 in the frequency ranges $30-78$ MHz and $120-168$ MHz. With complementary multi-frequency radio and X-ray data, we analyse the radio spectral properties of the halo and assess the connection between the non-thermal and thermal components of the ICM. We detected a third radio component (H3), which extends over the cluster volume on scales $\sim 2$ Mpc, embeds H1 and H2, and has a morphology that roughly follows the thermal ICM distribution. The radio spectral index is moderately steep in H1 ($\alpha=1.09\pm 0.02$) and H2 ($\alpha=1.15\pm 0.02$), but is steeper ($\alpha=1.57\pm 0.20$) in H3. The analysis of the thermal and non-thermal properties allowed us to discuss possible formation scenarios for each radio component. Turbulence from sloshing motions of low-entropy gas on different scales may be responsible for the origin of H1 and H2. We classified H3 as a giant ultra-steep spectrum radio halo, which could trace the residual activity from an old energetic merger and/or inefficient turbulent re-acceleration induced by ongoing minor mergers.

astro-ph.CO↗

arXiv2023

Requirements analysis for HPC\&HTC infrastructures integration in ESCAPE Science Analysis Platform

ESCAPE (European Science Cluster of Astronomy and Particle physics ESFRI research infrastructures) is a project to set up a cluster of ESFRI (European Strategy Forum on Research Infrastructures) facilities for astronomy, astroparticle and particle physics to face the challenges emerging through the modern multi-disciplinary data driven science. One of the main goal of ESCAPE is the building of ESAP (ESFRI Science Analysis Platform), a science platform for the analysis of open access data available through the EOSC (European Open Science Cloud) environment. ESAP will allow EOSC researchers to identify and stage existing data collections for analysis, share data, share and run scientific workflows. For many of the concerned ESFRIs and RIs, the data scales involved require significant computational resources (storage and compute) to support processing and analysis. The EOSC-ESFRI science platform therefore must implement appropriate interfaces to an underlying HPC (High Performance Computing) or HTC (High Throughput Computing) infrastructure to take advantage of it. This poster describes the analysis done to identify the main requirements for the implementation of the interfaces enabling the ESAP data access and computation resources integration in HPC and HTC computation infrastructures in terms of authentication and authorization policies, data management, workflow deployment and run.

astro-ph.IM↗

arXiv2023

The Planck clusters in the LOFAR sky. II. LoTSS-DR2: Recovering diffuse extended emission with LOFAR

Extended radio sources in the sky require a dense sampling of short baselines to be properly imaged by interferometers. This problem arises in many areas of radio astronomy, such as in the study of galaxy clusters, which may host Mpc-scale diffuse synchrotron sources in the form of radio halos. In clusters where no radio halos are detected, owing to intrinsic absence of emission or extrinsic (instrumental and/or observational) effects, it is possible to determine upper limits. We consider a sample of Planck galaxy clusters from the Second Data Release of the LOFAR Two Meter Sky Survey (LoTSS-DR2) where no radio halos are detected. We use this sample to test the capabilities of LOFAR to recover diffuse extended emission and derive upper limits. Through the injection technique, we simulate radio halos with various surface brightness profiles. We then predict the corresponding visibilities and image them along with the real visibilities. This method allows us to test the fraction of flux density losses owing to inadequate uv-coverage and obtain thresholds at which the mock emission becomes undetectable by visual inspection. The dense uv-coverage of LOFAR at short spacings allows to recover $\gtrsim90\%$ of the flux density of targets with sizes up to $\sim 15'$. We find a relation that provides upper limits based on the image noise and extent (in terms of number of beams) of the mock halo. This relation can be safely adopted to obtain upper limits without injecting when artifacts introduced by the subtraction of the discrete sources are negligible in the central region of the cluster. Otherwise, the injection process and visual inspection of the images are necessary to determine more reliable limits. Through these methods, we obtain upper limits for 75 clusters to be exploited in ongoing statistical studies.

astro-ph.CO↗

arXiv2022

Machine Learning to identify ICL and BCG in simulated galaxy clusters

Nowadays, Machine Learning techniques offer fast and efficient solutions for classification problems that would require intensive computational resources via traditional methods. We examine the use of a supervised Random Forest to classify stars in simulated galaxy clusters after subtracting the member galaxies. These dynamically different components are interpreted as the individual properties of the stars in the Brightest Cluster Galaxy (BCG) and IntraCluster Light (ICL). We employ matched stellar catalogues (built from the different dynamical properties of BCG and ICL) of 29 simulated clusters from the DIANOGA set to train and test the classifier. The input features are cluster mass, normalized particle cluster-centric distance, and rest-frame velocity. The model is found to correctly identify most of the stars, while the larger errors are exhibited at the BCG outskirt, where the differences between the physical properties of the two components are less obvious. We investigate the robustness of the classifier to numerical resolution, redshift dependence (up to $z=1$), and included astrophysical models. We claim that our classifier provides consistent results in simulations for $z<1$, at different resolution levels and with significantly different subgrid models. The phase-space structure is examined to assess whether the general properties of the stellar components are recovered: (i) the transition radius between BCG-dominated and ICL-dominated region is identified at $0.04$ \r200; (ii) the BCG outskirt ($> 0.1$ \r200) is significantly affected by uncertainties in the classification process. In conclusion, this work suggests the importance of employing Machine Learning to speed up a computationally expensive classification in simulations.

astro-ph.GA↗

arXiv2020

The DIANOGA simulations of galaxy clusters: characterizing star formation in proto-clusters

We studied the star formation rate (SFR) in cosmological hydrodynamical simulations of galaxy (proto-)clusters in the redshift range $0<z<4$, comparing them to recent observational studies; we also investigated the effect of varying the parameters of the star formation model on galaxy properties such as SFR, star-formation efficiency, and gas fraction. We analyze a set of zoom-in cosmological hydrodynamical simulations centred on twelve clusters. The simulations are carried out with the GADGET-3 TreePM/SPH code which includes various subgrid models to treat unresolved baryonic physics, including AGN feedback. Simulations do not reproduce the high values of SFR observed within protoclusters cores, where the values of SFR are underpredicted by a factor $\gtrsim 4$ both at $z\sim2$ and $z\sim 4$. The difference arises as simulations are unable to reproduce the observed starburst population and is worsened at $z\sim 2$ because simulations underpredict the normalization of the main sequence of star forming galaxies (i.e., the correlation between stellar mass and SFR) by a factor of $\sim 3$. As the low normalization of the main sequence seems to be driven by an underestimated gas fraction, it remains unclear whether numerical simulations miss starburst galaxies due to a too low predicted gas fractions or too low star formation efficiencies. Our results are stable against varying several parameters of the star formation subgrid model and do not depend on the details of the AGN feedback.

astro-ph.GA↗

arXiv2019

INAF Trieste Astronomical Observatory Information Technology Framework

INAF Trieste Astronomical Observatory (OATs) has a long tradition in information technology applied to Astronomical and Astrophysical use cases, particularly for what regards computing for data reduction, analysis and simulations; data and archives management; space missions data processing; design and software development for ground-based instruments. The ensemble of these activities, in the last years, pushed the need to acquire new computing resources and technologies and to deep competences in theirs management. In this paper we describe INAF-OATs computing centre technological stuff, our involvement in different EU Projects both in the path of building of EOSC, the European Open Science Cloud; in the design and prototyping of new Exascale supercomputers in Europe and the main research activities carried on using our computing centre.

astro-ph.IM↗

arXiv2019

Direct N-body application on low-power and energy-efficient parallel architectures

The aim of this work is to quantitatively evaluate the impact of computation on the energy consumption on ARM MPSoC platforms, exploiting CPUs, embedded GPUs and FPGAs. One of them possibly represents the future of High Performance Computing systems: a prototype of an Exascale supercomputer. Performance and energy measurements are made using a state-of-the-art direct $N$-body code from the astrophysical domain. We provide a comparison of the time-to-solution and energy delay product metrics, for different software configurations. We have shown that FPGA technologies can be used for application kernel acceleration and are emerging as a promising alternative to "traditional" technologies for HPC, which purely focus on peak-performance than on power-efficiency.

cs.PF↗

arXiv2019

Application of Google Cloud Platform in Astrophysics

The availability of new Cloud Platform offered by Google motivated us to propose nine Proof of Concepts (PoC) aiming to demonstrated and test the capabilities of the platform in the context of scientifically-driven tasks and requirements. We review the status of our initiative by illustrating 3 out of 9 successfully closed PoC that we implemented on Google Cloud Platform. In particular, we illustrate a cloud architecture for deployment of scientific software as microservice coupling Google Compute Engine with Docker and Pub/Sub to dispatch heavily parallel simulations. We detail also an experiment for HPC based simulation and workflow executions of data reduction pipelines (for the TNG-GIANO-B spectrograph) deployed on GCP. We compare and contrast our experience with on-site facilities comparing advantages and disadvantages both in terms of total cost of ownership and reached performances.

astro-ph.IM↗

arXiv2018

Cloud access to interoperable IVOA-compliant VOSpace storage

Handling, processing and archiving the huge amount of data produced by the new generation of experiments and instruments in Astronomy and Astrophysics are among the more exciting challenges to address in designing the future data management infrastructures and computing services. We investigated the feasibility of a data management and computation infrastructure, available world-wide, with the aim of merging the FAIR (Findable, Accessible, Interoperable, Reusable) data management provided by IVOA standards with the efficiency and reliability of a cloud approach. Our work involved the Canadian Advanced Network for Astronomy Research (CANFAR) infrastructure and the European EGI federated cloud (EFC). We designed and deployed a pilot data management and computation infrastructure that provides IVOA-compliant VOSpace storage resources and wide access to interoperable federated clouds. In this paper, we detail the main user requirements covered, the technical choices and the implemented solutions and we describe the resulting Hybrid cloud Worldwide infrastructure, its benefits and limitations.

astro-ph.IM↗

arXiv2017

Cosmological Simulations in Exascale Era

The architecture of Exascale computing facilities, which involves millions of heterogeneous processing units, will deeply impact on scientific applications. Future astrophysical HPC applications must be designed to make such computing systems exploitable. The ExaNeSt H2020 EU-funded project aims to design and develop an exascale ready prototype based on low-energy-consumption ARM64 cores and FPGA accelerators. We participate to the design of the platform and to the validation of the prototype with cosmological N-body and hydrodynamical codes suited to perform large-scale, high-resolution numerical simulations of cosmic structures formation and evolution. We discuss our activities on astrophysical applications to take advantage of the underlying architecture.

astro-ph.IM↗

arXiv2006

Bridging the Virtual Observatory and the GRID with the query element

The current generation of Grid infrastructures designed for production activity is strongly computing oriented and tuned on the needs of applications that requires intensive computations. Problems arise when trying to use such Grids to satisfy the sharing of data-oriented and service-oriented resources as happens in the IVOA community. We have designed, developed and implemented a Grid query element to access data source from an existing production Grid environment. We also enhanced the Grid middleware model (collective resources and sites) to manage Data Sources extending the Grid semantic. The query element and the modified grid Information System are able to connect the Grid environment to Virtual Observatory resources. A specialized query element is designed to work as Virtual Observatory resource in the Grid so than an Astronomer can access Virtual Observatory data using the IVOA standards.

astro-ph↗

arXiv2001

PINOCCHIO: pinpointing orbit-crossing collapsed hierarchical objects in a linear density field

PINOCCHIO (PINpointing Orbit-Crossing Collapsed Hierarchical Objects) is a new algorithm for identifying dark matter halos in a given numerical realisation of the linear density field in a hierarchical universe (Monaco et al. 2001). It is shown that Lagrangian perturbation theory, and in particular its ellipsoidal truncation, is able to predict accurately the collapse, in the orbit-crossing sense, of generic mass elements. Some points that have undergone orbit crossing are assigned to the network of filaments and sheets that connects the halos; it is demonstrated that this network resembles closely that found in N-body simulations. The code generates a catalogue of dark matter halos with known mass, position, velocity, merging history and angular momentum. It is shown that the predictions of the code are very accurate when compared with the results of large N-body simulations that cover a range of cosmological models, box sizes and numerical resolutions. The mass function is recovered with an accuracy of better than 10 per cent in number density for halos with at least 30-50 particles. A similar accuracy is reached in the estimate of the correlation length r_0. The good agreement is still valid on the object-by-object level, with 70-100 per cent of the objects with more than 50 particles in the simulations also identified by our algorithm. For these objects the masses are recovered with an error of 20-40 per cent, and positions and velocities with a root mean square error of ~1-2 Mpc (0.5-2 grid lengths) and ~100 km/s, respectively. The recovery of the angular momentum of halos is considerably noisier and accuracy at the statistical level is achieved only by introducing free parameters. The algorithm requires negligible computer time as compared with performing a numerical $N$-body simulation.

astro-ph↗

arXiv2001

PINOCCHIO and the hierarchical build-up of dark matter haloes

We study the ability of PINOCCHIO (PINpointing Orbit-Crossing Collapsed HIerarchical Objects) to predict the merging histories of dark matter (DM) haloes, comparing the PINOCCHIO predictions with the results of two large N-body simulations run from the same set of initial conditions. We focus our attention on quantities most relevant to galaxy formation and large-scale structure studies. PINOCCHIO is able to predict the statistics of merger trees with a typical accuracy of 20 per cent. Its validity extends to higher-order moments of the distribution of progenitors. The agreement is valid also at the object-by-object level, with 70-90 per cent of the progenitors cleanly recognised when the parent halo is cleanly recognised itself. Predictions are presented also for quantities that are usually not reproduced by semi-analytic codes, such as the two-point correlation function of the progenitors of massive haloes and the distribution of initial orbital parameters of merging haloes. For the accuracy of the prediction and for the facility with which merger histories are produced, PINOCCHIO provides a means to generate catalogues of DM haloes which is extremely competitive to large-scale N-body simulations, making it a suitable tool for galaxy formation and large-scale structure studies.

astro-ph↗

arXiv2001

Predicting the number, spatial distribution and merging history of dark matter haloes

We present a new algorithm (PINOCCHIO, PINpointing Orbit-Crossing Collapsed HIerarchical objects) to predict accurately the formation and evolution of individual dark matter haloes in a given realization of an initial linear density field. Compared with the halo population formed in a large (360^3 particles) collisionless simulation of a CDM universe, our method is able to predict to better than 10 per cent statistical quantities such as the mass function, two-point correlation function and progenitor mass function of the haloes. Masses of individual haloes are estimated accurately as well, with errors typically of order 30 per cent in the mass range well resolved by the numerical simulation. These results show that the hierarchical formation of dark matter haloes can be accurately predicted using local approximations to the dynamics when the correlations in the initial density field are properly taken into account. The approach allows one to automatically generate a large ensemble of accurate merging histories of haloes with complete knowledge of their spatial distribution. The construction of the full merger tree for a 256^3 realisation requires a few hours of CPU-time on a personal computer, orders of magnitude faster than the corresponding N-body simulation would take, and without needing any extensive post-processing. The technique can be efficiently used, for instance, for generating the input for galaxy formation modeling.

astro-ph↗