More simd/Neon functions

glm/detail/func_geometric_simd.inl

@@ -96,4 +96,48 @@ namespace detail
 }//namespace detail
 }//namespace glm
 
+#elif GLM_ARCH & GLM_ARCH_NEON_BIT
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_length<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& v)
+		{
+			return compute_dot<vec<4, float, Q>, float, true>::call(v, v);
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_distance<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& p0, vec<4, float, Q> const& p1)
+		{
+			return compute_length<4, float, Q, true>::call(p1 - p0);
+		}
+	};
+
+
+	template<qualifier Q>
+	struct compute_dot<vec<4, float, Q>, float, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& x, vec<4, float, Q> const& y)
+		{
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+			float32x4_t v = vmulq_f32(x.data, y.data);
+			v = vpaddq_f32(v, v);
+			v = vpaddq_f32(v, v);
+			return vgetq_lane_f32(v, 0);
+#else  // Armv7a with Neon
+			float32x4_t p = vmulq_f32(x.data, y.data);
+			float32x2_t v = vpadd_f32(vget_low_f32(p), vget_high_f32(p));
+			v = vpadd_f32(v, v);
+			return vget_lane_f32(v, 0);
+#endif
+		}
+	};
+}//namespace detail
+}//namespace glm
+
 #endif//GLM_ARCH & GLM_ARCH_SSE2_BIT