Added SSE2 protokernel to 4ic_deinterleave_8i_x2

kernels/volk/volk_4ic_deinterleave_8i_x2.h

@@ -60,4 +60,50 @@ static inline void volk_4ic_deinterleave_8i_x2_generic(int8_t* iBuffer,
 }
 
 #endif /* LV_HAVE_GENERIC */
+
+#ifdef LV_HAVE_SSE2
+#include <emmintrin.h>
+
+static inline void volk_4ic_deinterleave_8i_x2_a_sse2(int8_t* iBuffer,
+												   	   int8_t* qBuffer,
+													   const int8_t* complexVector,
+													   unsigned int num_points)
+{
+    // SSE2 algorithm was written by Andrey Semashev, licensed as CC-BY-SA
+    // https://stackoverflow.com/questions/63200053/deinterleve-vector-of-nibbles-using-simd
+	const __m128i mask = _mm_set1_epi32(0x0F0F0F0F);
+	const __m128i signed_max = _mm_set1_epi32(0x07070707);
+
+    unsigned int number = 0;
+    const unsigned int sixteenthPoints = num_points / 16;
+	for (; number < sixteenthPoints; number++)
+	{
+		// Load and deinterleave input half-bytes
+		__m128i input_even = _mm_loadu_si128(((const __m128i*)(complexVector)) + number);
+		__m128i input_odd  = _mm_srli_epi32(input_even, 4);
+
+		input_even = _mm_and_si128(input_even, mask);
+		input_odd  = _mm_and_si128(input_odd, mask);
+
+		// Get the sign bits
+		__m128i sign_even = _mm_cmpgt_epi8(input_even, signed_max);
+		__m128i sign_odd  = _mm_cmpgt_epi8(input_odd, signed_max);
+
+		// Combine sign bits with deinterleaved input
+		input_even = _mm_or_si128(input_even, _mm_andnot_si128(mask, sign_even));
+		input_odd  = _mm_or_si128(input_odd, _mm_andnot_si128(mask, sign_odd));
+
+		// Store the results
+		_mm_storeu_si128(((__m128i*)(iBuffer)) + number, input_even);
+		_mm_storeu_si128(((__m128i*)(qBuffer)) + number, input_odd);
+	}
+
+    number = sixteenthPoints * 16;
+    for (; number < num_points; number++) {
+        iBuffer[number] = complexVector[number] >> 4;
+        qBuffer[number] = (complexVector[number] << 4) >> 4;
+    }
+}
+
+#endif /* LV_HAVE_SSE2 */
 #endif /* INCLUDED_volk_4ic_deinterleave_8i_x2_u_H */