FL-MISR: fast large-scale multi-image super-resolution for computed tomography based on multi-GPU acceleration

Multi-image super-resolution (MISR) usually outperforms single-image super-resolution (SISR) under a proper inter-image alignment by explicitly exploiting the inter-image correlation. However, the large computational demand encumbers the deployment of MISR in practice. In this work, we propose a distributed optimization framework based on data parallelism for fast large-scale MISR using multi-GPU acceleration named FL-MISR. The scaled conjugate gradient (SCG) algorithm is applied to the distributed subfunctions and the local SCG variables are communicated to synchronize the convergence rate over multi-GPU systems towards a consistent convergence. Furthermore, an inner-outer border exchange scheme is performed to obviate the border effect between neighboring GPUs. The proposed FL-MISR is applied to the computed tomography (CT) system by super-resolving the projections acquired by subpixel detector shift. The SR reconstruction is performed on the fly during the CT acquisition such that no additional computation time is introduced. FL-MISR is extensively evaluated from different aspects and experimental results demonstrate that FL-MISR effectively improves the spatial resolution of CT systems in modulation transfer function (MTF) and visual perception. Comparing to a multi-core CPU implementation, FL-MISR achieves a more than 50$$\times$$ × speedup on an off-the-shelf 4-GPU system.

Adding GPU Acceleration to an Industrial CPU-Based Simulator, Development Strategy and Results

Running multi-million cell simulation problems in minutes has been a dream for reservoir engineers for decades. Today, with the advancement of Graphic Processing Unit (GPU), we have a real chance to make this dream a reality. Here we present our experience in the step-by-step transformation of a fully developed industrial CPU-based simulator into a fully functional GPU-based simulator. We also demonstrate significant accelerations achieved through the use of GPU technology. To achieve the best performance possible, we choose to use CUDA (NVIDIA GPU’s native language), and offload as much computations to GPU as possible. Our CUDA implementation covers all reservoir computes, which include property calculation, linearization, linear solver, etc. The well and Field Management still reside on CPU and need minor changes for their interaction with GPU-based reservoir. Importantly, there is no change to the nonlinear logic. The GPU and CPU parts are overlapped, fully utilizing the asynchronous nature of GPU operations. Each reservoir computation can be run in three modes, CPU_only (existing one), GPU_only, CPU followed by GPU. The latter is only used for result checking and debugging. In early 2019, we prototyped two reservoir linearization operations (mass accumulation and mass flux) in CUDA; both showed very strong runtime speed-up of several hundred times, 1 P100-GPU (NVIDIA) vs 1 POWER8NVL CPU core rated at 2.8 GHz (IBM). Encouraged by this success, we moved into linear solver development and managed to move the entire linear solver module into GPU. Again, strong speed-up of ~50 times was achieved (1 GPU vs 1 CPU). The focus for 2019 has been on standard Black-Oil cases. Our implementation was tested with multiple "million-cell range" models (SPE10 and other real field cases). In early 2020, we managed to put SPE10 fully on GPU, and finished the entire 2000 day time-stepping in ~35 sec with a single P100 card. After that our effort has switched to compositional AIM (Adaptive Implicit Method), with focus on compositional flash and AIM implementation for reservoir linearization and linear solver, both show early promising results. GPU-based reservoir simulation is a future trend for HPC. The development of a reservoir simulator is complex, multi-discipline and time-consuming work. Our paper demonstrates a clear strategy to add tremendous GPU acceleration into an existing CPU-based simulator. Our approach fully utilizes the strength of the existing CPU simulator and minimizes the GPU development effort. This paper is also the first publication targeting GPU acceleration for compositional AIM models.

Performance of a geometric deep learning pipeline for HL-LHC particle tracking

The Exa.TrkX project has applied geometric learning concepts such as metric learning and graph neural networks to HEP particle tracking. Exa.TrkX’s tracking pipeline groups detector measurements to form track candidates and filters them. The pipeline, originally developed using the TrackML dataset (a simulation of an LHC-inspired tracking detector), has been demonstrated on other detectors, including DUNE Liquid Argon TPC and CMS High-Granularity Calorimeter. This paper documents new developments needed to study the physics and computing performance of the Exa.TrkX pipeline on the full TrackML dataset, a first step towards validating the pipeline using ATLAS and CMS data. The pipeline achieves tracking efficiency and purity similar to production tracking algorithms. Crucially for future HEP applications, the pipeline benefits significantly from GPU acceleration, and its computational requirements scale close to linearly with the number of particles in the event.