AltiVec

AltiVec, officially AltiVec Technology and commonly referred to as Altivec or VMX, is a SIMD (single instruction, multiple data) extension to the PowerPC architecture. It was developed by Apple, IBM, and Motorola as part of the AIM alliance to provide data-parallel capabilities for multimedia, signal processing, and scientific workloads.

The extension introduces 32 128-bit vector registers, named V0 through V31, and a broad set of vector instructions that operate on packed data types such as 8-, 16-, 32-, and 64-bit integers and floating-point values. The instruction set covers arithmetic, logical, shift, comparison, and blend/select operations, along with vector load/store mechanisms. The AltiVec unit is integrated with the PowerPC core and accessed via the VMX facility.

History and usage notes: AltiVec appeared in the late 1990s, with the PowerPC 7400 (and the Macintosh G4 family) among the first implementations. Apple marketed it as Altivec, while IBM and Motorola used VMX as the underlying instruction set name. It was designed to accelerate multimedia, image and audio processing, and other data-parallel tasks.

Extensions and evolution: VSX is a later extension that augments the original VMX/AltiVec design by adding more vector and scalar capabilities for 64-bit PowerPC implementations. Compilers expose support for AltiVec/VMX and VSX through intrinsics and ISA-targeted options, allowing code to leverage the vector units where available.

Current status: On PowerPC-based systems, AltiVec/VMX remains part of the ISA and is supported by toolchains. In practice, its prominence declined as Apple migrated to Intel and then to ARM, but the technology influenced subsequent SIMD designs and remains a reference in high-performance vector processing.