Architectural choices for the Columbia 0.8 Teraflops machine
Igor V. Arsenin
Abstract
We discuss the hardware design choices made in our 16K-node 0.8 Teraflops supercomputer project, a machine architecture optimized for full QCD calculations. The efficiency of the conjugate gradient algorithm in terms of balance of floating-point operations, memory handling and utilization, and communication overhead is addressed. We also discuss the technological innovations and software tools that facilitate hardware design and what opportunities these give to the academic community.
Create a lesson
Related papers
Efficient Quantum Simulations of Yang-Mills theory with Maximal-tree Gauge
Tianyin Li, Ying-Ying Li, Xiaoyang Wang et al.
Physics-informed quantum algorithms for glueball-like excitations in a Z2 lattice gauge theory
Dan-Bo Zhang
Anomalous behavior of Wilson fermions in the presence of monopoles
Manuel Cortina, Rajamani Narayanan, Ray Romero
Direct lattice QCD calculation of the θ-induced CP-violating pion-nucleon coupling
Chuan-Yang Li, Jun Hua, Jian Liang et al.
Calculation of neutron electric dipole moment from Lattice QCD
Thomas Blum, Fangcheng He, Taku Izubuchi et al.
Exponential-in-Nc2 cost reduction of product-formula-based quantum simulations of quantum chromodynamics
Zohreh Davoudi, Jesse R. Stryker