summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--volk/apps/volk_profile.cc2
-rw-r--r--volk/include/volk/Makefile.am2
-rw-r--r--volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_a.h82
-rw-r--r--volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_u.h81
-rw-r--r--volk/lib/testqa.cc2
5 files changed, 169 insertions, 0 deletions
diff --git a/volk/apps/volk_profile.cc b/volk/apps/volk_profile.cc
index 712c32bce..6ba7f17bb 100644
--- a/volk/apps/volk_profile.cc
+++ b/volk/apps/volk_profile.cc
@@ -56,6 +56,8 @@ int main(int argc, char *argv[]) {
VOLK_PROFILE(volk_32fc_magnitude_squared_32f_u, 1e-4, 0, 204600, 1000, &results);
VOLK_PROFILE(volk_32fc_x2_multiply_32fc_a, 1e-4, 0, 204600, 1000, &results);
VOLK_PROFILE(volk_32fc_x2_multiply_32fc_u, 1e-4, 0, 204600, 1000, &results);
+ VOLK_PROFILE(volk_32fc_x2_multiply_conjugate_32fc_a, 1e-4, 0, 204600, 1000, &results);
+ VOLK_PROFILE(volk_32fc_x2_multiply_conjugate_32fc_u, 1e-4, 0, 204600, 1000, &results);
VOLK_PROFILE(volk_32f_s32f_convert_16i_a, 1, 32768, 204600, 10000, &results);
VOLK_PROFILE(volk_32f_s32f_convert_16i_u, 1, 32768, 204600, 10000, &results);
VOLK_PROFILE(volk_32f_s32f_convert_32i_a, 1, 2<<31, 204600, 10000, &results);
diff --git a/volk/include/volk/Makefile.am b/volk/include/volk/Makefile.am
index 20864efbe..d071f18f2 100644
--- a/volk/include/volk/Makefile.am
+++ b/volk/include/volk/Makefile.am
@@ -59,6 +59,8 @@ volkinclude_HEADERS = \
volk_32fc_32f_multiply_32fc_a.h \
volk_32fc_s32fc_multiply_32fc_a.h \
volk_32fc_s32fc_multiply_32fc_u.h \
+ volk_32fc_s32fc_multiply_conjugate_32fc_a.h \
+ volk_32fc_s32fc_multiply_conjugate_32fc_u.h \
volk_32fc_s32f_power_32fc_a.h \
volk_32f_s32f_calc_spectral_noise_floor_32f_a.h \
volk_32fc_s32f_atan2_32f_a.h \
diff --git a/volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_a.h b/volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_a.h
new file mode 100644
index 000000000..70476a8c7
--- /dev/null
+++ b/volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_a.h
@@ -0,0 +1,82 @@
+#ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
+#define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H
+
+#include <inttypes.h>
+#include <stdio.h>
+#include <volk/volk_complex.h>
+#include <float.h>
+
+#ifdef LV_HAVE_SSE3
+#include <pmmintrin.h>
+ /*!
+ \brief Multiplies vector a by the conjugate of vector b and stores the results in the third vector
+ \param cVector The vector where the results will be stored
+ \param aVector First vector to be multiplied
+ \param bVector Second vector that is conjugated before being multiplied
+ \param num_points The number of complex values in aVector and bVector to be multiplied together and stored into cVector
+ */
+static inline void volk_32fc_x2_multiply_conjugate_32fc_a_sse3(lv_32fc_t* cVector, const lv_32fc_t* aVector, const lv_32fc_t* bVector, unsigned int num_points){
+ unsigned int number = 0;
+ const unsigned int halfPoints = num_points / 2;
+
+ __m128 x, y, yl, yh, z, tmp1, tmp2;
+ lv_32fc_t* c = cVector;
+ const lv_32fc_t* a = aVector;
+ const lv_32fc_t* b = bVector;
+
+ __m128 conjugator = _mm_setr_ps(1, -1, 1, -1);
+
+ for(;number < halfPoints; number++){
+
+ x = _mm_load_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi
+ y = _mm_load_ps((float*)b); // Load the cr + ci, dr + di as cr,ci,dr,di
+
+ // FIXME: replace with xor for a faster implementation
+ y = _mm_mul_ps(y, conjugator); // conjugate y
+
+ yl = _mm_moveldup_ps(y); // Load yl with cr,cr,dr,dr
+ yh = _mm_movehdup_ps(y); // Load yh with ci,ci,di,di
+
+ tmp1 = _mm_mul_ps(x,yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
+
+ x = _mm_shuffle_ps(x,x,0xB1); // Re-arrange x to be ai,ar,bi,br
+
+ tmp2 = _mm_mul_ps(x,yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
+
+ z = _mm_addsub_ps(tmp1,tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
+
+ _mm_store_ps((float*)c,z); // Store the results back into the C container
+
+ a += 2;
+ b += 2;
+ c += 2;
+ }
+
+ if((num_points % 2) != 0) {
+ *c = (*a) * lv_conj(*b);
+ }
+}
+#endif /* LV_HAVE_SSE */
+
+#ifdef LV_HAVE_GENERIC
+ /*!
+ \brief Multiplies vector a by the conjugate of vector b and stores the results in the third vector
+ \param cVector The vector where the results will be stored
+ \param aVector First vector to be multiplied
+ \param bVector Second vector that is conjugated before being multiplied
+ \param num_points The number of complex values in aVector and bVector to be multiplied together and stored into cVector
+ */
+static inline void volk_32fc_x2_multiply_conjugate_32fc_a_generic(lv_32fc_t* cVector, const lv_32fc_t* aVector, const lv_32fc_t* bVector, unsigned int num_points){
+ lv_32fc_t* cPtr = cVector;
+ const lv_32fc_t* aPtr = aVector;
+ const lv_32fc_t* bPtr= bVector;
+ unsigned int number = 0;
+
+ for(number = 0; number < num_points; number++){
+ *cPtr++ = (*aPtr++) * lv_conj(*bPtr++);
+ }
+}
+#endif /* LV_HAVE_GENERIC */
+
+
+#endif /* INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_a_H */
diff --git a/volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_u.h b/volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_u.h
new file mode 100644
index 000000000..fbaa29c17
--- /dev/null
+++ b/volk/include/volk/volk_32fc_x2_multiply_conjugate_32fc_u.h
@@ -0,0 +1,81 @@
+#ifndef INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H
+#define INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H
+
+#include <inttypes.h>
+#include <stdio.h>
+#include <volk/volk_complex.h>
+#include <float.h>
+
+#ifdef LV_HAVE_SSE3
+#include <pmmintrin.h>
+ /*!
+ \brief Multiplies vector a by the conjugate of vector b and stores the results in the third vector
+ \param cVector The vector where the results will be stored
+ \param aVector First vector to be multiplied
+ \param bVector Second vector that is conjugated before being multiplied
+ \param num_points The number of complex values in aVector and bVector to be multiplied together and stored into cVector
+ */
+static inline void volk_32fc_x2_multiply_conjugate_32fc_u_sse3(lv_32fc_t* cVector, const lv_32fc_t* aVector, const lv_32fc_t* bVector, unsigned int num_points){
+ unsigned int number = 0;
+ const unsigned int halfPoints = num_points / 2;
+
+ __m128 x, y, yl, yh, z, tmp1, tmp2;
+ lv_32fc_t* c = cVector;
+ const lv_32fc_t* a = aVector;
+ const lv_32fc_t* b = bVector;
+
+ __m128 conjugator = _mm_set_ps(0, 0x80000000, 0, 0x80000000);
+
+ for(;number < halfPoints; number++){
+
+ x = _mm_loadu_ps((float*)a); // Load the ar + ai, br + bi as ar,ai,br,bi
+ y = _mm_loadu_ps((float*)b); // Load the cr + ci, dr + di as cr,ci,dr,di
+
+ y = _mm_xor_ps(y, conjugator); // conjugate y
+
+ yl = _mm_moveldup_ps(y); // Load yl with cr,cr,dr,dr
+ yh = _mm_movehdup_ps(y); // Load yh with ci,ci,di,di
+
+ tmp1 = _mm_mul_ps(x,yl); // tmp1 = ar*cr,ai*cr,br*dr,bi*dr
+
+ x = _mm_shuffle_ps(x,x,0xB1); // Re-arrange x to be ai,ar,bi,br
+
+ tmp2 = _mm_mul_ps(x,yh); // tmp2 = ai*ci,ar*ci,bi*di,br*di
+
+ z = _mm_addsub_ps(tmp1,tmp2); // ar*cr-ai*ci, ai*cr+ar*ci, br*dr-bi*di, bi*dr+br*di
+
+ _mm_storeu_ps((float*)c,z); // Store the results back into the C container
+
+ a += 2;
+ b += 2;
+ c += 2;
+ }
+
+ if((num_points % 2) != 0) {
+ *c = (*a) * lv_conj(*b);
+ }
+}
+#endif /* LV_HAVE_SSE */
+
+#ifdef LV_HAVE_GENERIC
+ /*!
+ \brief Multiplies vector a by the conjugate of vector b and stores the results in the third vector
+ \param cVector The vector where the results will be stored
+ \param aVector First vector to be multiplied
+ \param bVector Second vector that is conjugated before being multiplied
+ \param num_points The number of complex values in aVector and bVector to be multiplied together and stored into cVector
+ */
+static inline void volk_32fc_x2_multiply_conjugate_32fc_u_generic(lv_32fc_t* cVector, const lv_32fc_t* aVector, const lv_32fc_t* bVector, unsigned int num_points){
+ lv_32fc_t* cPtr = cVector;
+ const lv_32fc_t* aPtr = aVector;
+ const lv_32fc_t* bPtr= bVector;
+ unsigned int number = 0;
+
+ for(number = 0; number < num_points; number++){
+ *cPtr++ = (*aPtr++) * lv_conj(*bPtr++);
+ }
+}
+#endif /* LV_HAVE_GENERIC */
+
+
+#endif /* INCLUDED_volk_32fc_x2_multiply_conjugate_32fc_u_H */
diff --git a/volk/lib/testqa.cc b/volk/lib/testqa.cc
index b00ea0b64..fdd3d4853 100644
--- a/volk/lib/testqa.cc
+++ b/volk/lib/testqa.cc
@@ -91,6 +91,8 @@ VOLK_RUN_TESTS(volk_8i_s32f_convert_32f_a, 1e-4, 100, 20460, 1);
VOLK_RUN_TESTS(volk_8i_s32f_convert_32f_u, 1e-4, 100, 20460, 1);
VOLK_RUN_TESTS(volk_32fc_x2_multiply_32fc_a, 1e-4, 0, 20460, 1);
VOLK_RUN_TESTS(volk_32fc_x2_multiply_32fc_u, 1e-4, 0, 20460, 1);
+VOLK_RUN_TESTS(volk_32fc_x2_multiply_conjugate_32fc_a, 1e-4, 0, 20460, 1);
+VOLK_RUN_TESTS(volk_32fc_x2_multiply_conjugate_32fc_u, 1e-4, 0, 20460, 1);
VOLK_RUN_TESTS(volk_32f_x2_multiply_32f_a, 1e-4, 0, 20460, 1);
VOLK_RUN_TESTS(volk_32f_x2_multiply_32f_u, 1e-4, 0, 20460, 1);
VOLK_RUN_TESTS(volk_32fc_s32fc_multiply_32fc_a, 1e-4, 0, 20460, 1);
generated by cgit (git 2.34.1) at 2024-12-21 15:06:15 +0000