We address inefficiencies in task scheduling, memory management, and scalability in GPU-resident sparse LU factorization with a two-level approach of sequentially scheduled coarse-grained blocks containing multiple fine-grained blocks managed with a lightweight static scheduler enabling multi-stream parallelism. Additionally, we design an intelligent memory caching mechanism for the fine-grained scheduler, which retains frequently accessed data in GPU memory. To further enhance scalability, we introduce a distributed memory design that partitions the input matrix using a 1D block-cyclic distribution and optimizes inter-GPU communication via NVLink. The multi-GPU design reaches a computational throughput of 6.46 TFLOP/s on four A100 GPUs, demonstrating promising scalability. This is up to 7x speedup over the latest SuperLU_DIST with 3D communication, 94x speedup over PanguLU, 16x speedup over PasTiX, and 10x speedup over our own coarse-grained dynamic scheduling implementation while reaching up to 21% of the A100’s theoretical peak performance.

Multiplying two sparse matrices (SpGEMM) is a common computational primitive used in many areas including graph algorithms, bioinformatics, algebraic multigrid solvers, and randomized sketching. Distributed-memory parallel algorithms for SpGEMM have mainly focused on sparsity-oblivious approaches that use 2D and 3D partitioning. Sparsity-aware 1D algorithms can theoretically reduce communication by not fetching nonzeros of the sparse matrices that do not participate in the multiplication. Here, we present a distributed-memory 1D SpGEMM algorithm and implementation. It uses MPI RDMA operations to mitigate the cost of packing/unpacking submatrices for communication, and it uses a block fetching strategy to avoid excessive finegrained messaging. Our results show that our 1D implementation outperforms state-of-the-art 2D and 3D implementations within CombBLAS for many configurations, inputs, and use cases, while remaining conceptually simpler.

We exploit the high memory bandwidth of AI-customized Cerebras CS-2 systems for seismic processing. By leveraging low-rank matrix approximation, we fit memory-hungry seismic applications onto memory-austere SRAM wafer-scale hardware, thus addressing a challenge arising in many wave-equation-based algorithms that rely on Multi-Dimensional Convolution (MDC) operators. Exploiting sparsity inherent in seismic data in the frequency domain, we implement embarrassingly parallel tile low-rank matrix-vector multiplications (TLR-MVM), which account for most of the elapsed time in MDC operations, to successfully solve the Multi-Dimensional Deconvolution (MDD) inverse problem. By reducing memory footprint along with arithmetic complexity, we fit a standard seismic benchmark dataset into the small local memories of Cerebras processing elements. Deploying TLR-MVM execution onto 48 CS-2 systems in support of MDD gives a sustained memory bandwidth of 92.58PB/s on 35, 784, 000 processing elements, a significant milestone that highlights the capabilities of AI-customized architectures to enable a new generation of seismic algorithms that will empower multiple technologies of our low-carbon future.

Adaptive Optics (AO) is a technology that permits to measure and mitigate the distortion effects of atmospheric turbulence on optical beams. AO must operate in real-time by controlling thousands of actuators to shape the surface of deformable mirrors deployed on ground-based telescopes to compensate for these distortions. The command vectors that trigger how each individual actuator should act to bend a portion of the mirror are obtained from Matrix-Vector Multiplications (MVM). We identify and leverage the data sparsity structure of these control matrices coming from the MAVIS instruments for the European Southern Observatory's Very Large Telescope. We provide performance evaluation on x86 and accelerator-based systems. We present the impact of tile low-rank (TLR) matrix approximations on time-to-solution for the MVM and assess the produced image quality. We achieve performance improvement up to two orders of magnitude for TLR-MVM compared to regular dense MVM, while maintaining the image quality.