m128

m128 is a shorthand term used in computer programming and documentation to denote a 128-bit SIMD data type or register, most commonly in the context of x86 processors’ SSE (Streaming SIMD Extensions). In practice, m128 often refers to the C/C++ data type __m128, which represents a 128-bit SSE register and is used with compiler intrinsics to perform vectorized operations.

A 128-bit width can accommodate various element configurations depending on the instruction set and data type being operated on. Common interpretations include four 32-bit floating-point values (single-precision), two 64-bit floating-point values (double-precision), or other element sizes such as eight 16-bit integers. The actual interpretation is determined by the specific SSE instruction being used, for example, adding four floats with _mm_add_ps or adding two doubles with _mm_add_pd.

Programming with m128 typically involves intrinsics provided by headers such as immintrin.h or xmmintrin.h in C/C++. These intrinsics map directly to SSE instructions, enabling explicit vectorized operations like loading, storing, arithmetic, and comparisons on 128-bit vectors. For example, _mm_load_ps loads four 32-bit floats from memory into an __m128, while _mm_store_ps writes them back.

Historically, m128 references arise from the first generation of SSE, which introduced 128-bit XMM registers (XMM0–XMM15). SSE2 expanded capabilities to handle doubles and integers. While newer SIMD extensions (such as AVX) use wider registers, 128-bit m128 types remain common for compatibility and certain performance-tuned code paths.