This paper presents a portable, GPU-accelerated implementation of a QR-based singular value computation algorithm in Julia. The singular value decomposition (SVD) is a fundamental numerical tool in scientific computing and machine learning, providing optimal low-rank matrix approximations. Its importance has increased even more in large-scale machine learning pipelines, including large language models (LLMs), where it enables low-rank adaptation (LoRA). The implemented algorithm is based on the classic two-stage QR reduction, consisting of successive matrix reduction to band form and bidiagonal form. Our implementation leverages Julia’s multiple dispatch and metaprogramming capabilities, integrating with the GPUArrays and KernelAbstractions frameworks to provide a unified type and hardware-agnostic function. It supports diverse GPU architectures and data types, and is, to our knowledge, the first GPU-accelerated singular value implementation to support Apple Metal GPUs and half precision. Performance results on multiple GPU backends and data types demonstrate that portability does not require sacrificing performance: the unified function outperforms most linear algebra libraries (MAGMA, SLATE, rocSOLVER, oneMKL) for matrix sizes larger than 1024 × 1024, and achieves 80%-90% of the performance of cuSOLVER for large matrices.

The Fujitsu A64FX ARM-based processor is used in supercomputers such as Fugaku in Japan and Isambard 2 in the UK and provides an interesting combination of hardware features such as Scalable Vector Extension (SVE), and native support for reduced-precision floating-point arithmetic. The goal of this paper is to explore performance of the Julia programming language on the A64FX processor, with a particular focus on reduced precision. Here, we present a performance study on axpy to verify the compilation pipeline, demonstrating that Julia can match the performance of tuned libraries. Additionally, we investigate Message Passing Interface (MPI) scalability and throughput analysis on Fugaku showing next to no significant overheads of Julia of its MPI interface. To explore the usability of Julia to target various floating-point precisions, we present results of ShallowWaters. j1, a shallow water model that can be executed a various levels of precision. Even for such complex applications, Julia's type-flexible programming paradigm offers both, productivity and performance.

In this paper, we present a framework for moving compute and data between processing elements in a distributed heterogeneous system. The implementation of the framework is based on the LLVM compiler toolchain combined with the UCX communication framework. The framework can generate binary machine code or LLVM bitcode for multiple CPU architectures and move the code to remote machines while dynamically optimizing and linking the code on the target platform. The remotely injected code can recursively propagate itself to other remote machines or generate new code. The goal of this paper is threefold: (a) to present an ar-chitecture and implementation of the framework that provides essential infrastructure to program a new class of disaggregated systems wherein heterogeneous programming elements such as compute nodes and data processing units (DPUs) are distributed across the system, (b) to demonstrate how the framework can be integrated with modern, high-level programming languages such as Julia, and (c) to demonstrate and evaluate a new class of eXtended Remote Direct Memory Access (X-RDMA) communication operations that are enabled by this framework. To evaluate the capabilities of the framework, we used a cluster with Fujitsu CPUs and heterogeneous cluster with Intel CPUs and BlueField-2 DPUs interconnected using high-performance RDMA fabric. We demonstrated an X-RDMA pointer chase application that outperforms an RDMA GET-based implementation by 70% and is as fast as Active Messages, but does not require function predeployment on remote platforms.

Derivatives are key to numerous science, engineering, and machine learning applications. While existing tools generate derivatives of programs in a single language, modern parallel applications combine a set of frameworks and languages to leverage available performance and function in an evolving hardware landscape. We propose a scheme for differentiating arbitrary DAG-based parallelism that preserves scalability and efficiency, implemented into the LLVM-based Enzyme automatic differentiation framework. By integrating with a full-fledged compiler backend, Enzyme can differentiate numerous parallel frameworks and directly control code generation. Combined with its ability to differentiate any LLVM-based language, this flexibility permits Enzyme to leverage the compiler tool chain for parallel and differentiation-specitic optimizations. We differentiate nine distinct versions of the LULESH and miniBUDE applications, written in different programming languages (C++, Julia) and parallel frameworks (OpenMP, MPI, RAJA, Julia tasks, MPI.jl), demonstrating similar scalability to the original program. On benchmarks with 64 threads or nodes, we find a differentiation overhead of 3.4–6.8× on C++ and 5.4–12.5× on Julia.

Computing derivatives is key to many algorithms in scientific computing and machine learning such as optimization, uncertainty quantification, and stability analysis. Enzyme is a LLVM compiler plugin that performs reverse-mode automatic differentiation (AD) and thus generates high performance gradients of programs in languages including C/C++, Fortran, Julia, and Rust. Prior to this work, Enzyme and other AD tools were not capable of generating gradients of GPU kernels. Our paper presents a combination of novel techniques that make Enzyme the first fully automatic reversemode AD tool to generate gradients of GPU kernels. Since unlike other tools Enzyme performs automatic differentiation within a general-purpose compiler, we are able to introduce several novel GPU and AD-specific optimizations. To show the generality and efficiency of our approach, we compute gradients of five GPU-based HPC applications, executed on NVIDIA and AMD GPUs. All benchmarks run within an order of magnitude of the original program's execution time. Without GPU and AD-specific optimizations, gradients of GPU kernels either fail to run from a lack of resources or have infeasible overhead. Finally, we demonstrate that increasing the problem size by either increasing the number of threads or increasing the work per thread, does not substantially impact the overhead from differentiation.

Applying differentiable programming techniques and machine learning algorithms to foreign programs requires developers to either rewrite their code in a machine learning framework, or otherwise provide derivatives of the foreign code. This paper presents Enzyme, a high-performance automatic differentiation (AD) compiler plugin for the LLVM compiler framework capable of synthesizing gradients of statically analyzable programs expressed in the LLVM intermediate representation (IR). Enzyme synthesizes gradients for programs written in any language whose compiler targets LLVM IR including C, C++, Fortran, Julia, Rust, Swift, MLIR, etc., thereby providing native AD capabilities in these languages. Unlike traditional source-to-source and operator-overloading tools, Enzyme performs AD on optimized IR. On a machine-learning focused benchmark suite including Microsoft's ADBench, AD on optimized IR achieves a geometric mean speedup of 4.2 times over AD on IR before optimization allowing Enzyme to achieve state-of-the-art performance. Packaging Enzyme for PyTorch and TensorFlow provides convenient access to gradients of foreign code with state-of-the-art performance, enabling foreign code to be directly incorporated into existing machine learning workflows.

We introduce a spatial model of concentration dynamics that supports the emergence of spatiotemporal inhomogeneities that engage in metabolism-boundary co-construction. These configurations exhibit disintegration following some perturbations, and self-repair in response to others. We define robustness as a viable configuration's tendency to return to its prior configuration in response to perturbations, and plasticity as a viable configuration's tendency to change to other viable configurations. These properties are demonstrated and quantified in the model, allowing us to map a space of viable configurations and their possible transitions. Combining robustness and plasticity provides a measure of viability as the average expected survival time under ongoing perturbation, and allows us to measure how viability is affected as the configuration undergoes transitions. The framework introduced here is independent of the specific model we used, and is applicable for quantifying robustness, plasticity, and viability in any computational model of artificial life that demonstrates the conditions for viability that we promote.

We introduce a spatial model of autopoiesis that demonstrates boundary-metabolism co-construction, disintegration following harsh perturbations, and self-repair in response to lighter perturbations. The system’s continued maintenance is equated with robustness and quantitatively investigated. This investigation highlights several important operational dependencies for robustness including spatiality, identity, stability, and appropriate specification of perturbations in the types of environments to which an entity must be robust.