Abstracts
Please upload your slides.
Paolo Bientinesi: Pedagogical overview of tensor contractions
Overview of the field (existing software).
Devin Matthews: The Tensor-Based Library Instantiation Software (TBLIS)
Jutho Haegeman: Symmetric tensor network algorithms in Julia
I will provide an overview of the ecosystem of tensor network Julia packages that we have been developing under the umbrella of the QuantumKitHub organization. This development was started in the Quantum Group at Ghent University, but is gradually turning into an international collaboration. A key feature is the very strong support for symmetries throughout the different algorithms, whereas the more recent focus is on improving support for automatic differentiation and GPUs.
Katharine Hyatt: Support for tensor operations in Julia libraries
The Julia community has developed a variety of libraries for performing low- and high-level tensor operations and implementing algorithms. In this talk I'll summarize the state of Juliaworld for tensor library developers, including addressing automatic differentiation and GPU capabilities.
Edward Valeev: TiledArray: generic framework for efficient data-sparse tensor algebra on distributed-memory heterogeneous platforms
Andreas Irmler: High-Performance Tensor Contractions with Cyclops
This talk explores how large-scale distributed tensor contractions advance quantum chemistry research. After a brief example, it identifies the core challenge: efficiently distributing contractions across thousands of MPI ranks. The Cyclops Tensor Framework is presented as a solution, with a focus on its design, scalability, and performance for these demanding computations.
Kalman Szenes:The Density Matrix Renormalization Group in Quantum Chemistry
The density matrix renormalization group (DMRG), relying on the 1-dimensional matrix product state (MPS) tensor network, has established itself as the method of choice for large systems where exact diagonalization procedures are no longer viable. This talk provides an overview of the DMRG algorithm and highlights its applications in quantum chemistry.
Ryan Richard: TensorWrapper: Wishful thinking?
The NWChemEx (NWX) team is committed to relying on a series of domain-specific languages (DSLs) to manage the complexity of developing and maintaining a quantum chemistry code. The NWX architecture calls for one of those DSLs to be for expressing tensor operations. Moreover, the design calls for full encapsulation of the performance details. This is a tall order that no current tensor library fully supports. To that end, the NWX team has begun developing TensorWrapper.
Juraj Hasik: Introduction to high-level tensor interface, survey results
Alexander Heinecke: Pedagogical introduction to performance portability and tensor compilers
Alex Breuer:Tensor Compilation: Exploring Search Spaces
We present the Tiled Execution Intermediate Representation (TEIR), a compact IR for high-performance tensor operations that expresses computation as a composition of primitives over subtensors ("tiles"). TEIR separates what is executed from how it is scheduled through two records: TEIR-Primitives and TEIR-Schedule. The TEIR-Primitives record defines three primitives: first-access, main, and last-access. TEIR-Schedule assigns a per-axis execution policy and specifies axis extents and strides.
Jiajia Li: High-Performance Sparse Tensor Libraries
This talk will present my group’s research and software infrastructure for sparse tensor operations, covering both element- and block-wise sparsity on CPUs and GPUs.
Örs Legeza: Recent advances in tensor network state methods: a journey from mathematical aspects towards industrial perspectives
A brief overview of recent advances in tensor network state (TNS) methods are presented that have the potential to broaden their scope of application radically for strongly correlated quantum many body systems. Novel mathematical models for hybrid multiNode-multiGPU parallelization on high-performance computing (HPC) infrastructures will be discussed. Scaling analysis on NVIDIA DGX-A100 and DXG-H100 platforms reaching quarter petaflops performance on a single node will be presented. We also report cutting-edge performance results via mixed precision ab initio DMRG Density Matrix Renormalization Group (DMRG) electronic structure calculations, adapted for state-of-the-art NVIDIA Blackwell technology, utilizing the Ozaki scheme for emulating FP64 arithmetic through the use of fixed-point compute resources. Finally, we showcase recent results obtained on IBM superconducting quantum processor with up to 144 qubits, together with classical validation, using state-of-the-art tensor network simulations via a novel Basis Update Galerkin (BUG) method, establishing agreement between quantum and classical approaches. We close our presentation discussing future possibilities via utilization of Blackwell technology in tree-like TNS calculations opening new research directions in material sciences and beyond.
|
Loading... 
