Ivy Peng

Jacob Wahlgren, Andong Hu, Roger Pearce et al.

SmartNIC Data Processing Units (DPUs) offer a promising solution for saving high-end CPU resources by offloading tasks to programmable cores near the network interface. In this work, we explore the feasibility of SmartNIC DPUs in supporting an asynchronous communication model called "fire-and-forget", particularly its core message routing service. We design a communication offloading engine called Buddy that decouples communication tasks from the application process. Buddy runs flexibly on SmartNIC DPUs such as the Nvidia BlueField-3 DPU and generic x86 CPUs. Our evaluation results in five applications identify the memory-to-communication ratio as a key predictor of the offloading performance. Host-dominated workloads, such as Quicksilver and Sparse Matrix Transpose, achieved up to 1.55x speedup with communication offloaded to the DPU. We further identify a 625x increase in DRAM traffic due to the absence of Direct Cache Access support on the DPU, highlighting a critical need in future SmartNIC designs.

Frej Larssen, Ivy Peng, Stefano Markidis

Pulse-level simulators are the lowest-level, most widely used abstraction layer for studying how quantum hardware responds to control signals, but they can be built on Hamiltonian models with very different fidelity and cost. This raises the question: which level of physical abstraction is sufficient for a given simulation objective? We study this question for a flux-tunable two-qubit superconducting device with a fixed bus coupler by comparing three Hamiltonian descriptions of the same hardware: an effective two-level model, a three-mode Duffing model, and a circuit-based transmon model in the charge basis. Using a realistic parameter set, we evaluate these models on a common benchmark suite spanning flux-dependent spectra, extracted two-qubit interaction terms, driven single-qubit dynamics, CZ gate dynamics, leakage outside the computational subspace, and runtime. Across the tested flux range, the Duffing model follows the circuit-based reference more closely than the effective model for static spectra and reduced two-qubit quantities, while in driven benchmarks, the multilevel models reveal effects absent in the effective description. Overall, the results support a layered use of abstraction in pulse-level simulation: effective models for reduced analyses, Duffing models as a practical multilevel default, and circuit-based models for high-fidelity reference simulation or detailed leakage analysis.

Ruimin Shi, Maya Gokhale, Pei-Hung Lin et al.

The RISC-V Vector Extension~(RVV) is a cornerstone for supporting compute throughout in scientific and machine learning workloads. Yet compiler support and performance monitoring on real RVV~1.0 hardware are still evolving. In this work, we design a suite of assembly microbenchmarks to establish performance ceilings and calibrate performance counters on RVV hardware. Leveraging the assembly benchmarks, we find that predication overhead and stride load pose performance challenges that current compiler cost models do not yet fully address. Moreover, we present the first evaluation of GCC~15 and LLVM~21 autovectorization in HPC and ML proxy applications. GCC~15 outperforms LLVM~21 in four out of six applications. LLVM~21 only outperforms GCC~15 in SGEMM and DGEMM, driven by more aggressive instruction reduction confirmed through validated \texttt{perf} counters on the RVV hardware. We further show that the default LMUL selection in compilers performs close to the optimal. To study the RVV support for product-level application, we also evaluate the state-vector quantum simulator, Google's Qsim, with both manual RVV intrinsics and compiler auto-vectorization, revealing immaturity in current RVV compiler for complicated memory access pattern.

Marco Pasquale, Andong Hu, Luca Pennati et al.

As computational complexity in electromagnetics increases with frequency, full-wave solvers become computationally infeasible for electrically large problems. To address this limitation, we present a shooting and bouncing rays (SBR) method for efficiently modeling electromagnetic backscattering of metallic objects in the high-frequency regime. The method couples multi-reflection geometrical-optics ray transport with a physical optics surface integral discretized over ray tubes. To reduce the massive ray-surface intersection search space, we use a bounding volume hierarchy (BVH) and organize the computation as a trace-integrate pipeline. The ray tracing generates hit data, and the physical optics integral is evaluated over valid intersections only. Numerical accuracy is controlled through an incident-ray sampling rule that mitigates phase aliasing in the discretized physical optics integration. The method is accelerated on NVIDIA and AMD GPUs and parallelized with MPI. We validate against analytical Mie solutions for a perfectly electrically conducting (PEC) sphere and demonstrate applicability to a complex aircraft geometry for monostatic radar cross-section prediction.

Gabin Schieffer, Ruimin Shi, Jie Ren et al.

Advances in GPU compute throughput and memory capacity brings significant opportunities to a wide range of workloads. However, efficiently utilizing these resources remains challenging, particularly because diverse application characteristics may result in imbalanced utilization. Multi-Instance GPU (MIG) is a promising approach to improve utilization by partitioning GPU compute and memory resources into fixed-size slices with isolation. Yet, its effectiveness and limitations in supporting HPC workloads remain an open question. We present a comprehensive system-level characterization of different GPU sharing options using real-world scientific, AI, and data analytics applications, including NekRS, LAMMPS, Llama3, and Qiskit. Our analysis reveals that while GPU sharing via MIG can significantly reduce resource underutilization, and enable system-level improvements in throughput and energy, interference still occurs through shared resources, such as power throttling. Our performance-resource scaling results indicate that coarse-grained provisioning for tightly coupled compute and memory resources often mismatches application needs. To address this mismatch, we propose a memory-offloading scheme that leverages the cache-coherent Nvlink-C2C interconnect to bridge the gap between coarse-grained resource slices and reduce resource underutilization.

Luca Pennati, Andong Hu, Ivy Peng et al.

GROMACS is a de-facto standard for classical Molecular Dynamics (MD). The rise of AI-driven interatomic potentials that pursue near-quantum accuracy at MD throughput now poses a significant challenge: embedding neural-network inference into multi-GPU simulations retaining high-performance. In this work, we integrate the MLIP framework DeePMD-kit into GROMACS, enabling domain-decomposed, GPU-accelerated inference across multi-node systems. We extend the GROMACS NNPot interface with a DeePMD backend, and we introduce a domain decomposition layer decoupled from the main simulation. The inference is executed concurrently on all processes, with two MPI collectives used each step to broadcast coordinates and to aggregate and redistribute forces. We train an in-house DPA-1 model (1.6 M parameters) on a dataset of solvated protein fragments. We validate the implementation on a small protein system, then we benchmark the GROMACS-DeePMD integration with a 15,668 atom protein on NVIDIA A100 and AMD MI250x GPUs up to 32 devices. Strong-scaling efficiency reaches 66% at 16 devices and 40% at 32; weak-scaling efficiency is 80% to 16 devices and reaches 48% (MI250x) and 40% (A100) at 32 devices. Profiling with the ROCm System profiler shows that >90% of the wall time is spent in DeePMD inference, while MPI collectives contribute <10%, primarily since they act as a global synchronization point. The principal bottlenecks are the irreducible ghost-atom cost set by the cutoff radius, confirmed by a simple throughput model, and load imbalance across ranks. These results demonstrate that production MD with near ab initio fidelity is feasible at scale in GROMACS.

Ruimin Shi, Gabin Schieffer, Pei-Hung Lin et al.

ARM SVE and RISC-V RVV are emerging vector architectures in high-end processors that support vectorization of flexible vector length. In this work, we leverage an important workload for quantum computing, quantum state-vector simulations, to understand whether high-performance portability can be achieved in a vector-length agnostic (VLA) design. We propose a VLA design and optimization techniques critical for achieving high performance, including VLEN-adaptive memory layout adjustment, load buffering, fine-grained loop control, and gate fusion-based arithmetic intensity adaptation. We provide an implementation in Google's Qsim and evaluate five quantum circuits of up to 36 qubits on three ARM processors, including NVIDIA Grace, AWS Graviton3, and Fujitsu A64FX. By defining new metrics and PMU events to quantify vectorization activities, we draw generic insights for future VLA designs. Our single-source implementation of VLA quantum simulations achieves up to 4.5x speedup on A64FX, 2.5x speedup on Grace, and 1.5x speedup on Graviton.

Stefano Markidis, Jonah Ekelund, Luca Pennati et al.

Geomagnetic storms are large-scale disturbances of the Earth's magnetosphere driven by solar wind interactions, posing significant risks to space-based and ground-based infrastructure. The Disturbance Storm Time (Dst) index quantifies geomagnetic storm intensity by measuring global magnetic field variations. This study applies symbolic regression to derive data-driven equations describing the temporal evolution of the Dst index. We use historical data from the NASA OMNIweb database, including solar wind density, bulk velocity, convective electric field, dynamic pressure, and magnetic pressure. The PySR framework, an evolutionary algorithm-based symbolic regression library, is used to identify mathematical expressions linking dDst/dt to key solar wind. The resulting models include a hierarchy of complexity levels and enable a comparison with well-established empirical models such as the Burton-McPherron-Russell and O'Brien-McPherron models. The best-performing symbolic regression models demonstrate superior accuracy in most cases, particularly during moderate geomagnetic storms, while maintaining physical interpretability. Performance evaluation on historical storm events includes the 2003 Halloween Storm, the 2015 St. Patrick's Day Storm, and a 2017 moderate storm. The results provide interpretable, closed-form expressions that capture nonlinear dependencies and thresholding effects in Dst evolution.

Stefano Markidis, Ivy Peng, Artur Podobas et al.

Numerical simulations of plasma flows are crucial for advancing our understanding of microscopic processes that drive the global plasma dynamics in fusion devices, space, and astrophysical systems. Identifying and classifying particle trajectories allows us to determine specific on-going acceleration mechanisms, shedding light on essential plasma processes. Our overall goal is to provide a general workflow for exploring particle trajectory space and automatically classifying particle trajectories from plasma simulations in an unsupervised manner. We combine pre-processing techniques, such as Fast Fourier Transform (FFT), with Machine Learning methods, such as Principal Component Analysis (PCA), k-means clustering algorithms, and silhouette analysis. We demonstrate our workflow by classifying electron trajectories during magnetic reconnection problem. Our method successfully recovers existing results from previous literature without a priori knowledge of the underlying system. Our workflow can be applied to analyzing particle trajectories in different phenomena, from magnetic reconnection, shocks to magnetospheric flows. The workflow has no dependence on any physics model and can identify particle trajectories and acceleration mechanisms that were not detected before.