Platform-aware programming is the practice of writing computationally intensive software while explicitly exploiting features of the target execution platform. These features may include deep memory hierarchies, multicore processors, multiprocessor nodes, cluster interconnects, GPUs, FPGAs, AI accelerators, and other specialized hardware resources.

This approach has long been essential for applications whose success depends on extracting the highest possible performance from high-throughput and high-performance computing (HTC/HPC) systems. Traditionally, it has been associated with domains such as scientific computing, where developers routinely target large-scale parallel computing systems.

Today, however, platform-aware programming is no longer confined to traditional HPC. The rapid adoption of heterogeneous computing in Artificial Intelligence (AI), machine learning, and large-scale data analytics has made hardware specialization increasingly common. At the same time, the growing diversity of processors, GPUs, FPGAs, and other accelerators has significantly increased the complexity of modern computing platforms.

Although academia and industry have developed numerous programming models that seek to hide hardware differences behind unified interfaces, peak performance often still requires exploiting architecture-specific capabilities. In practice, achieving the best performance on heterogeneous systems frequently demands heterogeneous programming.

Despite this reality, relatively little attention has been given to programming abstractions that explicitly embrace hardware diversity instead of attempting to conceal it. Most existing approaches focus primarily on performance portability through generic interfaces, leaving programmers with limited support for expressing platform-specific optimizations in a modular and maintainable way.

The PlatformAwareProgramming organization hosts open-source projects dedicated to advancing platform-aware programming. Our goal is to develop programming interfaces, libraries, and abstractions that help developers manage the complexity of writing software for increasingly diverse computing platforms while preserving access to hardware-specific optimizations.

Our projects target a wide range of architectures, including multicore CPUs, GPUs, FPGAs, AI accelerators, and distributed-memory clusters, enabling programmers to express platform-specific assumptions in a structured, modular, and extensible manner.

Our first initiative is PlatformAware.jl, an experimental package that brings structured platform-aware programming to the Julia ecosystem. The project explores language abstractions that allow programmers to express platform assumptions explicitly while maintaining modularity and extensibility, taking advantage of Julia's multiple dispatch and metaprogramming features.

More recently, the organization has also become home to platform-aware, an experimental Rust crate that implements the concepts behind structured platform-aware programming in the Rust ecosystem. The Rust implementation investigates how they can be realized in a statically compiled systems programming language with fundamentally different design principles. By leveraging Rust's expressive type system, ownership model, traits, procedural macros, and compile-time metaprogramming facilities, the crate provides a framework for expressing platform assumptions in a structured, modular, extensible, and type-safe manner. It also serves as an experimental platform for research on language abstractions for heterogeneous and high-performance computing, including extensible platform models, platform-aware dispatch without an underlying support for multiple dispatch like in Julia, and compile-time specialization.

This organization is maintained by researchers and students of the High-Performance Computing (HPC) Research Group at the Federal University of Ceará (UFC). The projects hosted here stem from ongoing research conducted within the Graduate Program in Computer Science (MDCC) at UFC. The principal contact for the organization is Prof. Francisco Heron de Carvalho Junior, leader of the HPC Research Group and a faculty member of MDCC/UFC.

• MDCC: https://mdcc.ufc.br
• UFC: https://www.ufc.br