// C++20 SIMD TENSOR DOT-PRODUCT ENGINE [src/modules/core_simd/engine.cpp]
template <AlignedVectorType FloatType = float>
[[nodiscard]] FloatType compute_inner_product(std::span<const FloatType> lhs, 
                                              std::span<const FloatType> rhs) const noexcept {
    #if defined(__AVX512F__)
    __m512 sum512 = _mm512_setzero_ps();
    for (size_t i = 0; i + 16 <= lhs.size(); i += 16) {
        __m512 va = _mm512_loadu_ps(&lhs[i]);
        __m512 vb = _mm512_loadu_ps(&rhs[i]);
        sum512 = _mm512_fmadd_ps(va, vb, sum512); // Hardware Fused Multiply-Add
    }
    return _mm512_reduce_add_ps(sum512);
    #endif
}