Most embedded developers use #pragma pack(1) to create portable binary record formats. But pack(1) forces the compiler to byte-decompose every field access – even perfectly aligned fields – resulting in 7x more operations on RISC-V, ARM, and Xtensa. This post shows the problem, the fix (packed + aligned), and how to lint struct alignment as part of your development workflow.

Inter is a platform for designers, developers, and creators to explore the intersection of technology, design, and culture. Through articles, case studies, and interviews, Inter examines the impact of technology on society, creativity, and human experience. From UX/UI design trends and digital ethics to product innovation and creative storytelling, Inter provides  insights and inspiration for those shaping the future of digital experiences.

Interrupt

Using `#pragma pack(1)` on structs is common in embedded C for portable binary formats, but it forces the compiler to byte-decompose every field access — even perfectly aligned ones — resulting in up to 7x more instructions on RISC-V, ARM Cortex-M0, and Xtensa. The fix is combining `__attribute__((packed, aligned(4)))`: this preserves the no-padding layout while telling the compiler the struct base pointer is always aligned, enabling native single-instruction stores for aligned fields. The post also covers struct evolution risks (new fields silently landing at misaligned offsets), binary compatibility concerns with arrays of packed structs, and how to catch alignment regressions in CI using the `struct-lint` tool (DWARF-based) or `pahole`.

The Hidden Cost of Misalignment