Jan André Reuter

With the diverse space of accelerators in HPC, OpenMP target offloading has become a prevalent approach in accelerator programming. However, performance analysis has historically been challenging due to the distribution of relevant events across distinct tooling interfaces. OMPT enables the observation of these events through a standardized runtime interface, thereby facilitating the execution of OpenMP-level offload analysis. This paper extends the OMPT-based measurement support of Score-P to include support for target offloading. We outline the implementation approach and portability workarounds necessary to accommodate the discrepancies between current runtime and compiler implementations. For assessment, we investigate the overhead and compare OMPT-reported durations with vendor low-level interfaces utilizing microbenchmarks and SPEChpc 2021 benchmark workloads for the OpenMP runtimes provided by AMD ROCm and the HPE Cray Compiler Environment. The findings of the study suggest that OMPT-based offload analysis is a valuable tool for the evaluation of specific configurations. However, it is important to note that runtime-dependent limitations persist, particularly with regard to timestamp behavior and robustness. Still, the portability of the interface allows analysis across different runtime implementations, without the need for supporting multiple vendor-specific interfaces.

Myelinated axons (nerve fibers) efficiently transmit signals throughout the brain via action potentials. Multiple methods that are sensitive to axon orientations, from microscopy to magnetic resonance imaging, aim to reconstruct the brain’s structural connectome. As billions of nerve fibers traverse the brain with various possible geometries at each point, resolving fiber crossings is necessary to generate accurate structural connectivity maps. However, doing so with specificity is a challenging task because signals originating from oriented fibers can be influenced by brain (micro)structures unrelated to myelinated axons. X-ray scattering can specifically probe myelinated axons due to the periodicity of the myelin sheath, which yields distinct peaks in the scattering pattern. Here, we show that small-angle X-ray scattering (SAXS) can be used to detect myelinated, axon-specific fiber crossings. We first demonstrate the capability using strips of human corpus callosum to create artificial double- and triple-crossing fiber geometries, and we then apply the method in mouse, pig, vervet monkey, and human brains. Given its specificity, capability of 3-dimensional sampling and high resolution, SAXS can serve as a ground truth for validating MRI as well as microscopy-based methods. Statement of Significance: To study how the nerve fibers in our brain are interconnected, scientists need to visualize their trajectories, which often cross one another. Here, we show the unique capacity of small-angle X-ray scattering (SAXS) to study these fiber crossings without use of labelling, taking advantage of SAXS’s specificity to myelin - the insulating sheath that is wrapped around nerve fibers. We use SAXS to detect double and triple crossing fibers and unveil intricate crossings in mouse, pig, vervet monkey, and human brains. This non-destructive method can uncover complex fiber trajectories and validate other less specific imaging methods (e.g., MRI or microscopy), towards accurate mapping of neuronal connectivity in the animal and human brain. Graphical

Amunts, et al., 2011; M. Axer, Grässel, et al., 2011) struggle with the correct reconstruction of crossing nerve fibers. The recently developed technique Scattered Light Imaging (SLI) uses light in the optical range to explore the substructure of (crossing) nerve fibers in whole brain sections with micrometer resolution, providing crucial information such as the orientations of crossing nerve fibers in a measured image pixel (Menzel, Reuter, et al., 2020). The measurement principle was first introduced by Menzel, Axer, et al. (2020). Unstained histological brain sections are illuminated from different angles and the transmitted (scattered) light behind the sample is measured under normal incidence (see Figure 1a). Every pixel in the resulting SLI image series contains a light intensity profile (SLI profile I(ϕ), cf. Figure 1b-c) which is characteristic for the brain tissue structure at this point. The number and position of significant peaks in the SLI profiles can be used, for example, to determine the individual orientations of (crossing) nerve fibers in each image pixel.

Purpose The technique 3D polarized light imaging (3D-PLI) allows to reconstruct nerve fiber orientations of postmortem brains with ultra-high resolution. To better understand the physical principles behind 3D-PLI and improve the accuracy and reliability of the reconstructed fiber orientations, numerical simulations are employed which use synthetic nerve fiber models as input. As the generation of fiber models can be challenging and very time-consuming, we have developed the open source FAConstructor tool which enables a fast and efficient generation of synthetic fiber models for 3D-PLI simulations. Methods The program was developed as an interactive tool, allowing the user to define fiber pathways with interpolation methods or parametric functions and providing visual feedback. Results Performance tests showed that most processes scale almost linearly with the amount of fiber points in FAConstructor. Fiber models consisting of < 1.6 million data points retain a frame rate of more than 30 frames per second, which guarantees a stable and fluent workflow. The applicability of FAConstructor was demonstrated on a well-defined fiber model (Fiber Cup phantom) for two different simulation approaches, reproducing effects known from 3D-PLI measurements. Conclusion We have implemented a user-friendly and efficient tool that enables an interactive and fast generation of synthetic nerve fiber configurations for 3D-PLI simulations. Already existing fiber models can easily be modified, allowing to simulate many different fiber models in a reasonable amount of time.