Thomas Pasquali

Characterizing the Scalability and Performance of Large-Scale AI Training Under Multi-Tenancy

Jacopo Raffi, Thomas Pasquali, Lorenzo Piarulli, Filippo Spiga et autres

Characterising AI workload performance on modern HPC systems requires understanding both their scalability in isolation and their behaviour under concurrent execution. However, the interplay among parallelisation strategies, network congestion, compute capability, and interconnect technologies remains poorly understood. This work investigates the performance …

DLNetBenchSC26

Jacopo Raffi, Thomas Pasquali, Loredana Piarulli, Filippo Spiga et autres

Artifact for SC26 paper "Characterizing the Scalability and Performance of Large-Scale AI Training Under Multi-Tenancy". Link Github: https://github.com/HicrestLaboratory/DLNetBenchSC26

it, gb, de (code pays fourni par la source)

DLNetBenchSC26

Jacopo Raffi, Thomas Pasquali, Loredana Piarulli, Filippo Spiga et autres

Artifact for SC26 paper "Characterizing the Scalability and Performance of Large-Scale AI Training Under Multi-Tenancy". Link Github: https://github.com/HicrestLaboratory/DLNetBenchSC26

it, gb, de (code pays fourni par la source)

Communication-Avoiding SpGEMM via Trident Partitioning on Hierarchical GPU Interconnect

Julian Bellavita, Lorenzo Pichetti, Thomas Pasquali, Flavio Vella et autres

The multiplication of two sparse matrices, known as SpGEMM, is a key kernel in scientific computing and large-scale data analytics, underpinning graph algorithms, machine learning, simulations, and computational biology, where sparsity is often highly unstructured. The unstructured sparsity makes achieving high performance …

us, it (code pays fourni par la source)

NET4EXA: Toward Ethernet-native interconnects for exascale HPC and AI systems

Matteo Turisini, John Gliksberg, Corentin Beaulieu, Mohamed Benazouz et autres

it, fr, gr (code pays fourni par la source)

Communication-Avoiding SpGEMM via Trident Partitioning on Hierarchical GPU Interconnects

Julian Bellavita, Lorenzo Pichetti, Thomas Pasquali, Flavio Vella et autres

The multiplication of two sparse matrices, known as SpGEMM, is a key kernel in scientific computing and large-scale data analytics, underpinning graph algorithms, machine learning, simulations, and computational biology, where sparsity is often highly unstructured. The unstructured sparsity makes achieving high performance …

Cache-optimized BFS on multi-core CPUs

Salvatore Domenico Andaloro, Thomas Pasquali, Flavio Vella

Breadth-First Search (BFS) performance on shared-memory systems is often limited by irregular memory access and cache inefficiencies. This work presents two optimizations for BFS graph traversal: a bitmap-based algorithm designed for small-diameter graphs and MergedCSR, a graph storage format that improves cache …

it (code pays fourni par la source)

Popcorn: Accelerating Kernel K-means on GPUs through Sparse Linear Algebra

Julian Bellavita, Thomas Pasquali, Laura Río-Martín, Flavio Vella et autres

K-means is a popular clustering algorithm with significant applications in numerous scientific and engineering areas. One drawback of K-means is its inability to identify non-linearly separable clusters, which may lead to inaccurate solutions in certain cases. Kernel K-means is a variant of …

it, us (code pays fourni par la source)