Abstract MLExchange is a machine learning (ML) operations platform providing web user-interfaces (UIs) for data visualization and analysis pipelines at synchrotron facilities. Among these UIs is the segmentation app which helps synchrotron users utilize ML algorithms to automatically segment high-resolution scientific images with minimal manual annotation effort. In this work, we share code optimizations that significantly speed up the segmentation inference workflow of large data in short time. By optimizing the sequence of CPU-GPU data transfers and introducing CPU parallelization to key operations, we improve the per-device, per-image frame computational efficiency and observe close to 3 $$\times$$ × speedup over the original segmentation inference workflow run time when utilizing a single GPU. Further adaptations enabling multi-GPU inference yield more than 40 $$\times$$ × speedup with 100 GPUs compared to the optimized single GPU inference workflow. This acceleration of the segmentation inference workflow will provide MLExchange users with easy access to segmentation results with little wait time.

Synchrotron-based soft X-ray ptychography has enabled the reconstruction of both the phase and attenuation projections of samples relevant to the physical and biological sciences. The phase projection images typically have higher fidelity and hence are used for tomographic reconstruction. In practice, three-dimensional tomographic reconstruction can be challenging because the measurements may have outliers, a fluctuating background and may be restricted to a limited angular range of sample rotations. Thus, conventional reconstruction algorithms such as filtered back projection can result in reconstructions with strong artifacts. In this paper, we present a robust model-based iterative reconstruction algorithm for X-ray ptychography-based phase tomography. Our method casts the reconstruction as a regularized inverse problem, involving a novel data fitting term that accounts for noise, the fluctuating background as well as outliers, combined with an image model term that enforces regularity on the volume to be reconstructed. We use a majorization-minimization strategy to find a minimum of the formulated cost function. Reconstructions on a simulated as well as a real dataset show that it is possible to acquire high-quality phase reconstructions compared to the typically used filtered-back projection algorithm as well as conventional regularized inversion approaches.

In the field of nanoparticle material science, X-ray scattering techniques are widely used for characterization of macromolecules and particle systems (ordered, partially-ordered or custom) based on their structural properties at the micro- and nano-scales. Numerous applications utilize these, including design and fabrication of energy-relevant nanodevices such as photovoltaic and energy storage devices. Due to its size, analysis of raw data obtained through present ultra-fast light beamlines and X-ray scattering detectors has been a primary bottleneck in such characterization processes. To address this hurdle, we are developing high-performance parallel algorithms and codes for analysis of X-ray scattering data for several of the scattering methods, such as the Small Angle X-ray Scattering (SAXS), which we talk about in this paper. As an inverse modeling problem, structural fitting of the raw data obtained through SAXS experiments is a method used for extracting meaningful information on the structural properties of materials. Such fitting processes involve a large number of variable parameters and, hence, require a large amount of computational power. In this paper, we focus on this problem and present a high-performance and scalable parallel solution based on the Reverse Monte Carlo simulation algorithm, on highly-parallel systems such as clusters of multicore CPUs and graphics processors. We have implemented and optimized our algorithm on generic multi-core CPUs as well as the Nvidia GPU architectures with C++ and CUDA. We also present detailed performance results and computational analysis of our code.

Although present X-ray scattering techniques can provide tremendous information on the nano-structural properties of materials that are valuable in the design and fabrication of energy-relevant nano-devices, a primary challenge remains in the analyses of such data. In this paper we describe a high-performance, flexible, and scalable Grazing Incidence Small Angle X-ray Scattering simulation algorithm and codes that we have developed on multi-core/CPU and many-core/GPU clusters. We discuss in detail our implementation, optimization and performance on these platforms. Our results show speedups of ~125x on a Fermi-GPU and ~20x on a Cray-XE6 24-core node, compared to a sequential CPU code, with near linear scaling on multi-node clusters. To our knowledge, this is the first GISAXS simulation code that is flexible to compute scattered light intensities in all spatial directions allowing full reconstruction of GISAXS patterns for any complex structures and with highresolutions while reducing simulation times from months to minutes.