[Commit-gnuradio] [gnuradio] 13/22: volk: add neon versions for 32i bitwise operators

[git]

This is an automated email from the git hooks/post-receive script.

jcorgan pushed a commit to branch master
in repository gnuradio.

commit 34670fd911d4f491d5b6c6140b703619283d1432
Author: Nathan West <address@hidden>
Date:   Mon Oct 20 19:07:29 2014 -0500

    volk: add neon versions for 32i bitwise operators
---
 volk/kernels/volk/volk_32i_x2_and_32i.h | 34 +++++++++++++++++++++++++++++++++
 volk/kernels/volk/volk_32i_x2_or_32i.h  | 34 +++++++++++++++++++++++++++++++++
 2 files changed, 68 insertions(+)

diff --git a/volk/kernels/volk/volk_32i_x2_and_32i.h 
b/volk/kernels/volk/volk_32i_x2_and_32i.h
index b33a60e..c138540 100644
--- a/volk/kernels/volk/volk_32i_x2_and_32i.h
+++ b/volk/kernels/volk/volk_32i_x2_and_32i.h
@@ -65,6 +65,40 @@ static inline void volk_32i_x2_and_32i_a_sse(int32_t* 
cVector, const int32_t* aV
 }
 #endif /* LV_HAVE_SSE */
 
+#ifdef LV_HAVE_NEON
+#include <arm_neon.h>
+/*!
+  \brief Ands the two input vectors and store their results in the third vector
+  \param cVector The vector where the results will be stored
+  \param aVector One of the vectors
+  \param bVector One of the vectors
+  \param num_points The number of values in aVector and bVector to be anded 
together and stored into cVector
+*/
+static inline void volk_32i_x2_and_32i_neon(int32_t* cVector, const int32_t* 
aVector, const int32_t* bVector, unsigned int num_points){
+    int32_t* cPtr = cVector;
+    const int32_t* aPtr = aVector;
+    const int32_t* bPtr=  bVector;
+    unsigned int number = 0;
+    unsigned int quarter_points = num_points / 4;
+
+    int32x4_t a_val, b_val, c_val;
+
+    for(number = 0; number < quarter_points; number++){
+        a_val = vld1q_s32(aPtr);
+        b_val = vld1q_s32(bPtr);
+        c_val = vandq_s32(a_val, b_val);
+        vst1q_s32(cPtr, c_val);
+        aPtr += 4;
+        bPtr += 4;
+        cPtr += 4;
+    }
+
+    for(number = quarter_points * 4; number < num_points; number++){
+      *cPtr++ = (*aPtr++) & (*bPtr++);
+    }
+}
+#endif /* LV_HAVE_NEON */
+
 #ifdef LV_HAVE_GENERIC
 /*!
   \brief Ands the two input vectors and store their results in the third vector
diff --git a/volk/kernels/volk/volk_32i_x2_or_32i.h 
b/volk/kernels/volk/volk_32i_x2_or_32i.h
index a8556a3..544a71c 100644
--- a/volk/kernels/volk/volk_32i_x2_or_32i.h
+++ b/volk/kernels/volk/volk_32i_x2_or_32i.h
@@ -65,6 +65,40 @@ static inline void volk_32i_x2_or_32i_a_sse(int32_t* 
cVector, const int32_t* aVe
 }
 #endif /* LV_HAVE_SSE */
 
+#ifdef LV_HAVE_NEON
+#include <arm_neon.h>
+/*!
+  \brief Ands the two input vectors and store their results in the third vector
+  \param cVector The vector where the results will be stored
+  \param aVector One of the vectors
+  \param bVector One of the vectors
+  \param num_points The number of values in aVector and bVector to be anded 
together and stored into cVector
+*/
+static inline void volk_32i_x2_or_32i_neon(int32_t* cVector, const int32_t* 
aVector, const int32_t* bVector, unsigned int num_points){
+    int32_t* cPtr = cVector;
+    const int32_t* aPtr = aVector;
+    const int32_t* bPtr=  bVector;
+    unsigned int number = 0;
+    unsigned int quarter_points = num_points / 4;
+
+    int32x4_t a_val, b_val, c_val;
+
+    for(number = 0; number < quarter_points; number++){
+        a_val = vld1q_s32(aPtr);
+        b_val = vld1q_s32(bPtr);
+        c_val = vorrq_s32(a_val, b_val);
+        vst1q_s32(cPtr, c_val);
+        aPtr += 4;
+        bPtr += 4;
+        cPtr += 4;
+    }
+
+    for(number = quarter_points * 4; number < num_points; number++){
+      *cPtr++ = (*aPtr++) | (*bPtr++);
+    }
+}
+#endif /* LV_HAVE_NEON */
+
 #ifdef LV_HAVE_GENERIC
 /*!
   \brief Ors the two input vectors and store their results in the third vector