Memory

Moving data between memory and SIMD registers efficiently. The difference between fast and slow SIMD code is usually in the memory access patterns, not the arithmetic.

1. Load & Store — Unaligned, aligned, partial, streaming
2. Gather & Scatter — Non-contiguous access and prefetch hints
3. Interleaved Data — deinterleave_4ch, interleave_4ch for RGBA and similar
4. Chunked Processing — Processing large arrays in SIMD-sized chunks, alignment, performance