Related papers: A computational system for lattice QCD with overla…
To unleash the potential of quantum computers, noise effects on qubits' performance must be carefully managed. The decoders responsible for diagnosing noise-induced computational errors must use resources efficiently to enable scaling to…
Modern interconnects often have programmable processors in the network interface that can be utilized to offload communication processing from host CPU. In this paper, we explore different schemes to support collective operations at the…
The rapid development of programmable network devices and the widespread use of machine learning (ML) in networking have facilitated efficient research into intelligent data plane (IDP). Offloading ML to programmable data plane (PDP)…
CPUs are critical for LLM serving due to their availability, cost efficiency, and edge applicability. However, efficient CPU serving is hindered by conflicting prefill/decode resource demands under non-disaggregated deployment…
We calculate the 2-loop partition function of QCD on the lattice, using the Wilson formulation for gluons and the overlap-Dirac operator for fermions. Direct by-products of our result are the 2-loop free energy and average plaquette. Our…
This paper proposes a composite inner-product computation unit based on left-to-right (LR) arithmetic for the acceleration of convolution neural networks (CNN) on hardware. The efficacy of the proposed L2R-CIPU method has been shown on the…
We present a new set of QCD codes in both message passing and data parallel versions. The message passing package used is PARMACS, although other packages may be used. Data parallel software is written in High Performance fortran, an…
The current status of United States projects pursuing Teraflops-scale computing resources for lattice field theory is discussed. Two projects are in existence at this time: the Multidisciplinary Teraflops Project, incorporating the…
Specialized hardware like application-specific integrated circuits (ASICs) remains the primary accelerator type for cryptographic kernels based on large integer arithmetic. Prior work has shown that commodity and server-class GPUs can…
Modern graphics hardware is designed for highly parallel numerical tasks and provides significant cost and performance benefits. Graphics hardware vendors are now making available development tools to support general purpose high…
We present results of a quenched QCD simulation with overlap fermions on a lattice of volume V = 16^3X32 at beta=6.0, which corresponds approximatively to a lattice cutoff of 2 GeV and an extension of 1.4 fm. From the two-point correlation…
Convolutional neural networks (CNNs) require high throughput hardware accelerators for real time applications owing to their huge computational cost. Most traditional CNN accelerators rely on single core, linear processing elements (PEs) in…
We study physics at temperatures just above the QCD phase transition (Tc) using chiral (overlap) Fermions in the quenched approximation of lattice QCD. Exact zero modes of the overlap Dirac operator are localized and their frequency of…
Using barebone PC components and NIC's, we construct a linux cluster which has 2-dimensional mesh structure. This cluster has smaller footprint, is less expensive, and use less power compared to conventional linux cluster. Here, we report…
We propose Shuttling-based Distributed Quantum Computing (SDQC), a hybrid architecture that combines the strengths of physical qubit shuttling and distributed quantum computing to enable scalable trapped-ion quantum computing. SDQC performs…
From the overlap lattice quark propagator calculated in the Landau gauge, we determine the quark chiral condensate by fitting operator product expansion formulas to the lattice data. The quark propagators are computed on domain wall fermion…
LINAC 4 is a normal conducting H- linac proposed at CERN to provide a higher proton flux to the CERN accelerator chain. It should replace the existing LINAC 2 as injector to the Proton Synchrotron Booster and can also operate in the future…
TCP/IP network stack is irreplaceable for Web services in datacenter front-end servers, and the demand for which is growing rapidly for emerging high concurrency network service applications (including Internet, Internet of Things, mobile…
Large deep learning models have demonstrated strong ability to solve many tasks across a wide range of applications. Those large models typically require training and inference to be distributed. Tensor parallelism is a common technique…
We have developed a new version of the high-performance J\"ulich universal quantum computer simulator (JUQCS-50) that leverages key features of the GH200 superchips as used in the JUPITER supercomputer, enabling simulations of a 50-qubit…