m128d

m128d is a 128‑bit vector data type used in x86 SIMD programming, primarily with the SSE2 instruction set. It is designed to hold two double-precision floating-point values, enabling parallel operations on two doubles in a single instruction.

In practice, the 128 bits are interpreted as two 64‑bit lanes, each storing one double. The type is commonly exposed in C/C++ headers as __m128d (the canonical SSE2 intrinsic type), though some documentation or bindings may refer to it as m128d. Access to this type and its operations is typically provided via intrinsics defined in headers such as emmintrin.h or immintrin.h, depending on the compiler.

Common operations on m128d include arithmetic, comparisons, and data movement. Typical intrinsics cover addition, subtraction, multiplication, and division of corresponding lanes, for example _mm_add_pd, _mm_sub_pd, _mm_mul_pd, and _mm_div_pd. Memory access intrinsics allow loading and storing two doubles from memory, such as _mm_load_pd and _mm_store_pd, or broadcasting a scalar to both lanes with _mm_set1_pd. Alignment plays a role: aligned loads/stores require 16-byte alignment, while unaligned variants like _mm_loadu_pd enable safe use with unaligned memory.

Availability and requirements: m128d relies on SSE2 support, which is available on most modern x86 CPUs. Compiler support requires enabling SSE2 features (for example, -msse2 in GCC/Clang or /arch:SSE2 in MSVC). The actual name and headers may vary by compiler, but the underlying concept remains a 128-bit vector containing two doubles for parallel numeric computation.