mirrors/gnss-sdr
1
0
Fork 0
mirror of https://github.com/gnss-sdr/gnss-sdr synced 2025-10-31 07:13:03 +00:00
Code Issues Releases Wiki Activity

Merge branch 'next' of https://github.com/gnss-sdr/gnss-sdr into next

Browse Source
This commit is contained in:
Carles Fernandez
2016-01-31 14:00:12 +01:00
parent 8be7d9e2a0 f4875012df
commit 3f6e819df5
2 changed files with 35 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

60
src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_s32fc_x2_rotator_16ic.h
View File

@@ -79,11 +79,11 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_generic(lv_16sc_t* ou
static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* outVector, const lv_16sc_t* inVector, const lv_32fc_t phase_inc, lv_32fc_t* phase, unsigned int num_points) static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* outVector, const lv_16sc_t* inVector, const lv_32fc_t phase_inc, lv_32fc_t* phase, unsigned int num_points)
{ {
const unsigned int sse_iters = num_points / 4; const unsigned int sse_iters = num_points / 4;
__m128 a,b, two_phase_acc_reg,two_phase_inc_reg; __m128 a, b, two_phase_acc_reg, two_phase_inc_reg;
__m128i c1,c2,result; __m128i c1, c2, result;
__attribute__((aligned(16))) lv_32fc_t two_phase_inc[2]; __attribute__((aligned(16))) lv_32fc_t two_phase_inc[2];
two_phase_inc[0] = phase_inc*phase_inc; two_phase_inc[0] = phase_inc * phase_inc;
two_phase_inc[1] = phase_inc*phase_inc; two_phase_inc[1] = phase_inc * phase_inc;
two_phase_inc_reg = _mm_load_ps((float*) two_phase_inc); two_phase_inc_reg = _mm_load_ps((float*) two_phase_inc);
__attribute__((aligned(16))) lv_32fc_t two_phase_acc[2]; __attribute__((aligned(16))) lv_32fc_t two_phase_acc[2];
two_phase_acc[0] = (*phase); two_phase_acc[0] = (*phase);
@@ -95,6 +95,8 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
lv_16sc_t* _out = outVector; lv_16sc_t* _out = outVector;
__m128 yl, yh, tmp1, tmp2, tmp3; __m128 yl, yh, tmp1, tmp2, tmp3;
lv_16sc_t tmp16;
lv_32fc_t tmp32;
for(unsigned int number = 0; number < sse_iters; number++) for(unsigned int number = 0; number < sse_iters; number++)
{ {
@@ -105,7 +107,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
b=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di b = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
c1 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic c1 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic
//complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg //complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg
@@ -114,7 +116,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(two_phase_inc_reg, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(two_phase_inc_reg, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
tmp3 = _mm_shuffle_ps(two_phase_inc_reg, two_phase_inc_reg, 0xB1); // Re-arrange x to be ai,ar,bi,br tmp3 = _mm_shuffle_ps(two_phase_inc_reg, two_phase_inc_reg, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(tmp3, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(tmp3, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
two_phase_acc_reg=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di two_phase_acc_reg = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
//next two samples //next two samples
_in += 2; _in += 2;
@@ -125,7 +127,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
b=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di b = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
c2 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic c2 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic
//complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg //complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg
@@ -134,7 +136,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(two_phase_inc_reg, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(two_phase_inc_reg, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
tmp3 = _mm_shuffle_ps(two_phase_inc_reg, two_phase_inc_reg, 0xB1); // Re-arrange x to be ai,ar,bi,br tmp3 = _mm_shuffle_ps(two_phase_inc_reg, two_phase_inc_reg, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(tmp3, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(tmp3, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
two_phase_acc_reg=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di two_phase_acc_reg = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
// store four output samples // store four output samples
result = _mm_packs_epi32(c1, c2);// convert from 32ic to 16ic result = _mm_packs_epi32(c1, c2);// convert from 32ic to 16ic
@@ -146,9 +148,8 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
} }
_mm_storeu_ps((float*)two_phase_acc, two_phase_acc_reg); _mm_storeu_ps((float*)two_phase_acc, two_phase_acc_reg);
(*phase) = lv_cmake(two_phase_acc[0], two_phase_acc[0]) * phase_inc; (*phase) = lv_cmake(two_phase_acc[0], two_phase_acc[0]);
lv_16sc_t tmp16;
lv_32fc_t tmp32;
for (unsigned int i = sse_iters * 4; i < num_points; ++i) for (unsigned int i = sse_iters * 4; i < num_points; ++i)
{ {
tmp16 = *_in++; tmp16 = *_in++;
@@ -166,15 +167,15 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_a_sse3(lv_16sc_t* out
static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* outVector, const lv_16sc_t* inVector, const lv_32fc_t phase_inc, lv_32fc_t* phase, unsigned int num_points) static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* outVector, const lv_16sc_t* inVector, const lv_32fc_t phase_inc, lv_32fc_t* phase, unsigned int num_points)
{ {
const unsigned int sse_iters = num_points / 4; const unsigned int sse_iters = num_points / 4;
__m128 a,b, two_phase_acc_reg,two_phase_inc_reg; __m128 a, b, two_phase_acc_reg, two_phase_inc_reg;
__m128i c1,c2,result; __m128i c1, c2, result;
__attribute__((aligned(16))) lv_32fc_t two_phase_inc[2]; __attribute__((aligned(16))) lv_32fc_t two_phase_inc[2];
two_phase_inc[0] = phase_inc*phase_inc; two_phase_inc[0] = phase_inc * phase_inc;
two_phase_inc[1] = phase_inc*phase_inc; two_phase_inc[1] = phase_inc * phase_inc;
two_phase_inc_reg = _mm_load_ps((float*) two_phase_inc); two_phase_inc_reg = _mm_load_ps((float*) two_phase_inc);
__attribute__((aligned(16))) lv_32fc_t two_phase_acc[2]; __attribute__((aligned(16))) lv_32fc_t two_phase_acc[2];
two_phase_acc[0] = (*phase); two_phase_acc[0] = (*phase);
two_phase_acc[1] = (*phase)*phase_inc; two_phase_acc[1] = (*phase) * phase_inc;
two_phase_acc_reg = _mm_load_ps((float*) two_phase_acc); two_phase_acc_reg = _mm_load_ps((float*) two_phase_acc);
const lv_16sc_t* _in = inVector; const lv_16sc_t* _in = inVector;
@@ -182,6 +183,8 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* out
lv_16sc_t* _out = outVector; lv_16sc_t* _out = outVector;
__m128 yl, yh, tmp1, tmp2, tmp3; __m128 yl, yh, tmp1, tmp2, tmp3;
lv_16sc_t tmp16;
lv_32fc_t tmp32;
for(unsigned int number = 0; number < sse_iters; number++) for(unsigned int number = 0; number < sse_iters; number++)
{ {
@@ -192,7 +195,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
b=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di b = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
c1 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic c1 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic
//complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg //complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg
@@ -201,7 +204,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(two_phase_inc_reg, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(two_phase_inc_reg, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
tmp3 = _mm_shuffle_ps(two_phase_inc_reg, two_phase_inc_reg, 0xB1); // Re-arrange x to be ai,ar,bi,br tmp3 = _mm_shuffle_ps(two_phase_inc_reg, two_phase_inc_reg, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(tmp3, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(tmp3, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
two_phase_acc_reg=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di two_phase_acc_reg = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
//next two samples //next two samples
_in += 2; _in += 2;
@@ -212,7 +215,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* out
tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr tmp1 = _mm_mul_ps(a, yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br a = _mm_shuffle_ps(a, a, 0xB1); // Re-arrange x to be ai,ar,bi,br
tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di tmp2 = _mm_mul_ps(a, yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
b=_mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di b = _mm_addsub_ps(tmp1, tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
c2 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic c2 = _mm_cvtps_epi32(b); // convert from 32fc to 32ic
//complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg //complex 32fc multiplication two_phase_acc_reg=two_phase_acc_reg*two_phase_inc_reg
@@ -233,9 +236,8 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_u_sse3(lv_16sc_t* out
} }
_mm_storeu_ps((float*)two_phase_acc, two_phase_acc_reg); _mm_storeu_ps((float*)two_phase_acc, two_phase_acc_reg);
(*phase) = lv_cmake(two_phase_acc[0], two_phase_acc[0]) * phase_inc; (*phase) = lv_cmake(two_phase_acc[0], two_phase_acc[0]);
lv_16sc_t tmp16;
lv_32fc_t tmp32;
for (unsigned int i = sse_iters * 4; i < num_points; ++i) for (unsigned int i = sse_iters * 4; i < num_points; ++i)
{ {
tmp16 = *_in++; tmp16 = *_in++;
@@ -261,8 +263,8 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_neon(lv_16sc_t* outVe
lv_16sc_t* _out = outVector; lv_16sc_t* _out = outVector;
lv_32fc_t ___phase4 = phase_inc * phase_inc * phase_inc * phase_inc; lv_32fc_t ___phase4 = phase_inc * phase_inc * phase_inc * phase_inc;
float32_t __phase4_real[4] = { lv_creal(___phase4), lv_creal(___phase4), lv_creal(___phase4), lv_creal(___phase4) }; __VOLK_ATTR_ALIGNED(16) float32_t __phase4_real[4] = { lv_creal(___phase4), lv_creal(___phase4), lv_creal(___phase4), lv_creal(___phase4) };
float32_t __phase4_imag[4] = { lv_cimag(___phase4), lv_cimag(___phase4), lv_cimag(___phase4), lv_cimag(___phase4) }; __VOLK_ATTR_ALIGNED(16) float32_t __phase4_imag[4] = { lv_cimag(___phase4), lv_cimag(___phase4), lv_cimag(___phase4), lv_cimag(___phase4) };
float32x4_t _phase4_real = vld1q_f32(__phase4_real); float32x4_t _phase4_real = vld1q_f32(__phase4_real);
float32x4_t _phase4_imag = vld1q_f32(__phase4_imag); float32x4_t _phase4_imag = vld1q_f32(__phase4_imag);
@@ -271,8 +273,8 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_neon(lv_16sc_t* outVe
lv_32fc_t phase3 = phase2 * phase_inc; lv_32fc_t phase3 = phase2 * phase_inc;
lv_32fc_t phase4 = phase3 * phase_inc; lv_32fc_t phase4 = phase3 * phase_inc;
float32_t __phase_real[4] = { lv_creal((*phase)), lv_creal(phase2), lv_creal(phase3), lv_creal(phase4) }; __VOLK_ATTR_ALIGNED(16) float32_t __phase_real[4] = { lv_creal((*phase)), lv_creal(phase2), lv_creal(phase3), lv_creal(phase4) };
float32_t __phase_imag[4] = { lv_cimag((*phase)), lv_cimag(phase2), lv_cimag(phase3), lv_cimag(phase4) }; __VOLK_ATTR_ALIGNED(16) float32_t __phase_imag[4] = { lv_cimag((*phase)), lv_cimag(phase2), lv_cimag(phase3), lv_cimag(phase4) };
float32x4_t _phase_real = vld1q_f32(__phase_real); float32x4_t _phase_real = vld1q_f32(__phase_real);
float32x4_t _phase_imag = vld1q_f32(__phase_imag); float32x4_t _phase_imag = vld1q_f32(__phase_imag);
@@ -288,7 +290,9 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_neon(lv_16sc_t* outVe
for(; i < neon_iters; ++i) for(; i < neon_iters; ++i)
{ {
/* load 4 complex numbers (int 16 bits each component) */ /* load 4 complex numbers (int 16 bits each component) */
tmp16 = vld2_s16((int16_t*)_in); _in += 4; tmp16 = vld2_s16((int16_t*)_in);
__builtin_prefetch(_in + 8);
_in += 4;
/* promote them to int 32 bits */ /* promote them to int 32 bits */
tmp32i.val[0] = vmovl_s16(tmp16.val[0]); tmp32i.val[0] = vmovl_s16(tmp16.val[0]);
@@ -339,7 +343,7 @@ static inline void volk_gnsssdr_16ic_s32fc_x2_rotator_16ic_neon(lv_16sc_t* outVe
vst1q_f32((float32_t*)__phase_real, _phase_real); vst1q_f32((float32_t*)__phase_real, _phase_real);
vst1q_f32((float32_t*)__phase_imag, _phase_imag); vst1q_f32((float32_t*)__phase_imag, _phase_imag);
(*phase) = lv_cmake((float32_t)__phase_real[3], (float32_t)__phase_imag[3]) * phase_inc; (*phase) = lv_cmake((float32_t)__phase_real[0], (float32_t)__phase_imag[0]);
} }
for(i = 0; i < neon_iters % 4; ++i) for(i = 0; i < neon_iters % 4; ++i)
{ {

8
src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_x2_dot_prod_16ic.h
View File

@@ -77,6 +77,7 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_a_sse2(lv_16sc_t* out, con
if (sse_iters > 0) if (sse_iters > 0)
{ {
__m128i a,b,c, c_sr, mask_imag, mask_real, real, imag, imag1,imag2, b_sl, a_sl, realcacc, imagcacc, result; __m128i a,b,c, c_sr, mask_imag, mask_real, real, imag, imag1,imag2, b_sl, a_sl, realcacc, imagcacc, result;
__VOLK_ATTR_ALIGNED(16) lv_16sc_t dotProductVector[4];
realcacc = _mm_setzero_si128(); realcacc = _mm_setzero_si128();
imagcacc = _mm_setzero_si128(); imagcacc = _mm_setzero_si128();
@@ -116,8 +117,6 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_a_sse2(lv_16sc_t* out, con
result = _mm_or_si128 (realcacc, imagcacc); result = _mm_or_si128 (realcacc, imagcacc);
__VOLK_ATTR_ALIGNED(16) lv_16sc_t dotProductVector[4];
_mm_store_si128((__m128i*)dotProductVector,result); // Store the results back into the dot product vector _mm_store_si128((__m128i*)dotProductVector,result); // Store the results back into the dot product vector
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
@@ -154,6 +153,7 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_u_sse2(lv_16sc_t* out, con
if (sse_iters > 0) if (sse_iters > 0)
{ {
__m128i a,b,c, c_sr, mask_imag, mask_real, real, imag, imag1,imag2, b_sl, a_sl, realcacc, imagcacc, result; __m128i a,b,c, c_sr, mask_imag, mask_real, real, imag, imag1,imag2, b_sl, a_sl, realcacc, imagcacc, result;
__VOLK_ATTR_ALIGNED(16) lv_16sc_t dotProductVector[4];
realcacc = _mm_setzero_si128(); realcacc = _mm_setzero_si128();
imagcacc = _mm_setzero_si128(); imagcacc = _mm_setzero_si128();
@@ -193,8 +193,6 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_u_sse2(lv_16sc_t* out, con
result = _mm_or_si128 (realcacc, imagcacc); result = _mm_or_si128 (realcacc, imagcacc);
__VOLK_ATTR_ALIGNED(16) lv_16sc_t dotProductVector[4];
_mm_storeu_si128((__m128i*)dotProductVector,result); // Store the results back into the dot product vector _mm_storeu_si128((__m128i*)dotProductVector,result); // Store the results back into the dot product vector
for (int i = 0; i < 4; ++i) for (int i = 0; i < 4; ++i)
@@ -227,7 +225,7 @@ static inline void volk_gnsssdr_16ic_x2_dot_prod_16ic_neon(lv_16sc_t* out, const
// 2nd lane holds the imaginary part // 2nd lane holds the imaginary part
int16x4x2_t a_val, b_val, c_val, accumulator; int16x4x2_t a_val, b_val, c_val, accumulator;
int16x4x2_t tmp_real, tmp_imag; int16x4x2_t tmp_real, tmp_imag;
lv_16sc_t accum_result[4]; __VOLK_ATTR_ALIGNED(16) lv_16sc_t accum_result[4];
accumulator.val[0] = vdup_n_s16(0); accumulator.val[0] = vdup_n_s16(0);
accumulator.val[1] = vdup_n_s16(0); accumulator.val[1] = vdup_n_s16(0);