28 #ifndef INCLUDE_VOLK_VOLK_AVX_INTRINSICS_H_ 29 #define INCLUDE_VOLK_VOLK_AVX_INTRINSICS_H_ 30 #include <immintrin.h> 35 __m256 yl, yh, tmp1, tmp2;
36 yl = _mm256_moveldup_ps(y);
37 yh = _mm256_movehdup_ps(y);
38 tmp1 = _mm256_mul_ps(x, yl);
39 x = _mm256_shuffle_ps(x, x, 0xB1);
40 tmp2 = _mm256_mul_ps(x, yh);
41 return _mm256_addsub_ps(tmp1, tmp2);
46 const __m256 conjugator = _mm256_setr_ps(0, -0.f, 0, -0.f, 0, -0.f, 0, -0.f);
47 return _mm256_xor_ps(x, conjugator);
58 __m256 complex1, complex2;
59 cplxValue1 = _mm256_mul_ps(cplxValue1, cplxValue1);
60 cplxValue2 = _mm256_mul_ps(cplxValue2, cplxValue2);
61 complex1 = _mm256_permute2f128_ps(cplxValue1, cplxValue2, 0x20);
62 complex2 = _mm256_permute2f128_ps(cplxValue1, cplxValue2, 0x31);
63 return _mm256_hadd_ps(complex1, complex2);
73 __m256 sign_mask_dummy = _mm256_setzero_ps();
74 const __m128i zeros = _mm_set1_epi8(0x00);
75 const __m128i sign_extract = _mm_set1_epi8(0x80);
76 const __m128i shuffle_mask0 = _mm_setr_epi8(0xff, 0xff, 0xff, 0x00, 0xff, 0xff, 0xff, 0x01, 0xff, 0xff, 0xff, 0x02, 0xff, 0xff, 0xff, 0x03);
77 const __m128i shuffle_mask1 = _mm_setr_epi8(0xff, 0xff, 0xff, 0x04, 0xff, 0xff, 0xff, 0x05, 0xff, 0xff, 0xff, 0x06, 0xff, 0xff, 0xff, 0x07);
79 fbits = _mm_cmpgt_epi8(fbits, zeros);
80 fbits = _mm_and_si128(fbits, sign_extract);
81 __m128i sign_bits0 = _mm_shuffle_epi8(fbits, shuffle_mask0);
82 __m128i sign_bits1 = _mm_shuffle_epi8(fbits, shuffle_mask1);
84 __m256 sign_mask = _mm256_insertf128_ps(sign_mask_dummy, _mm_castsi128_ps(sign_bits0), 0x0);
85 return _mm256_insertf128_ps(sign_mask, _mm_castsi128_ps(sign_bits1), 0x1);
94 __m256 part0 = _mm256_permute2f128_ps(src0, src1, 0x20);
95 __m256 part1 = _mm256_permute2f128_ps(src0, src1, 0x31);
96 *llr0 = _mm256_shuffle_ps(part0, part1, 0x88);
97 *llr1 = _mm256_shuffle_ps(part0, part1, 0xdd);
102 const __m256 sign_mask = _mm256_set1_ps(-0.0f);
103 const __m256 abs_mask = _mm256_andnot_ps(sign_mask, _mm256_castsi256_ps(_mm256_set1_epi8(0xff)));
109 __m256 sign = _mm256_xor_ps(_mm256_and_ps(llr0, sign_mask), _mm256_and_ps(llr1, sign_mask));
110 __m256 dst = _mm256_min_ps(_mm256_and_ps(llr0, abs_mask), _mm256_and_ps(llr1, abs_mask));
111 return _mm256_or_ps(dst, sign);
123 llr0 = _mm256_xor_ps(llr0, sign_mask);
124 __m256 dst = _mm256_add_ps(llr0, llr1);
static void _mm256_polar_deinterleave(__m256 *llr0, __m256 *llr1, __m256 src0, __m256 src1)
Definition: volk_avx_intrinsics.h:92
static __m256 _mm256_complexmul_ps(__m256 x, __m256 y)
Definition: volk_avx_intrinsics.h:33
static __m256 _mm256_complexconjugatemul_ps(__m256 x, __m256 y)
Definition: volk_avx_intrinsics.h:51
static __m256 _mm256_conjugate_ps(__m256 x)
Definition: volk_avx_intrinsics.h:45
static __m256 _mm256_magnitude_ps(__m256 cplxValue1, __m256 cplxValue2)
Definition: volk_avx_intrinsics.h:67
static __m256 _mm256_polar_minsum_llrs(__m256 src0, __m256 src1)
Definition: volk_avx_intrinsics.h:101
static __m256 _mm256_magnitudesquared_ps(__m256 cplxValue1, __m256 cplxValue2)
Definition: volk_avx_intrinsics.h:57
static __m256 _mm256_polar_sign_mask(__m128i fbits)
Definition: volk_avx_intrinsics.h:72
static __m256 _mm256_polar_fsign_add_llrs(__m256 src0, __m256 src1, __m128i fbits)
Definition: volk_avx_intrinsics.h:115