diff options
Diffstat (limited to 'volk')
| -rw-r--r-- | volk/include/volk/volk_32f_s32f_convert_16i_a.h | 15 | ||||
| -rw-r--r-- | volk/include/volk/volk_32f_s32f_convert_16i_u.h | 15 |
2 files changed, 16 insertions, 14 deletions
diff --git a/volk/include/volk/volk_32f_s32f_convert_16i_a.h b/volk/include/volk/volk_32f_s32f_convert_16i_a.h index 10c921b08..a24959678 100644 --- a/volk/include/volk/volk_32f_s32f_convert_16i_a.h +++ b/volk/include/volk/volk_32f_s32f_convert_16i_a.h @@ -4,6 +4,7 @@ #include <volk/volk_common.h> #include <inttypes.h> #include <stdio.h> +#include <math.h> #ifdef LV_HAVE_SSE2 #include <emmintrin.h> @@ -57,7 +58,7 @@ static inline void volk_32f_s32f_convert_16i_a_sse2(int16_t* outputVector, const r = max_val; else if(r < min_val) r = min_val; - outputVector[number] = (int16_t)(r); + outputVector[number] = (int16_t)rintf(r); } } #endif /* LV_HAVE_SSE2 */ @@ -98,10 +99,10 @@ static inline void volk_32f_s32f_convert_16i_a_sse(int16_t* outputVector, const ret = _mm_max_ps(_mm_min_ps(_mm_mul_ps(ret, vScalar), vmax_val), vmin_val); _mm_store_ps(outputFloatBuffer, ret); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[0]); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[1]); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[2]); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[3]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[0]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[1]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[2]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[3]); } number = quarterPoints * 4; @@ -111,7 +112,7 @@ static inline void volk_32f_s32f_convert_16i_a_sse(int16_t* outputVector, const r = max_val; else if(r < min_val) r = min_val; - outputVector[number] = (int16_t)(r); + outputVector[number] = (int16_t)rintf(r); } } #endif /* LV_HAVE_SSE */ @@ -138,7 +139,7 @@ static inline void volk_32f_s32f_convert_16i_a_generic(int16_t* outputVector, co r = min_val; else if(r > max_val) r = max_val; - *outputVectorPtr++ = (int16_t)(r); + *outputVectorPtr++ = (int16_t)rintf(r); } } #endif /* LV_HAVE_GENERIC */ diff --git a/volk/include/volk/volk_32f_s32f_convert_16i_u.h b/volk/include/volk/volk_32f_s32f_convert_16i_u.h index f339a7d10..f58158041 100644 --- a/volk/include/volk/volk_32f_s32f_convert_16i_u.h +++ b/volk/include/volk/volk_32f_s32f_convert_16i_u.h @@ -3,6 +3,7 @@ #include <inttypes.h> #include <stdio.h> +#include <math.h> #ifdef LV_HAVE_SSE2 #include <emmintrin.h> @@ -57,7 +58,7 @@ static inline void volk_32f_s32f_convert_16i_u_sse2(int16_t* outputVector, const r = max_val; else if(r < min_val) r = min_val; - outputVector[number] = (int16_t)(r); + outputVector[number] = (int16_t)rintf(r); } } #endif /* LV_HAVE_SSE2 */ @@ -99,10 +100,10 @@ static inline void volk_32f_s32f_convert_16i_u_sse(int16_t* outputVector, const ret = _mm_max_ps(_mm_min_ps(_mm_mul_ps(ret, vScalar), vmax_val), vmin_val); _mm_store_ps(outputFloatBuffer, ret); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[0]); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[1]); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[2]); - *outputVectorPtr++ = (int16_t)(outputFloatBuffer[3]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[0]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[1]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[2]); + *outputVectorPtr++ = (int16_t)rintf(outputFloatBuffer[3]); } number = quarterPoints * 4; @@ -112,7 +113,7 @@ static inline void volk_32f_s32f_convert_16i_u_sse(int16_t* outputVector, const r = max_val; else if(r < min_val) r = min_val; - outputVector[number] = (int16_t)(r); + outputVector[number] = (int16_t)rintf(r); } } #endif /* LV_HAVE_SSE */ @@ -140,7 +141,7 @@ static inline void volk_32f_s32f_convert_16i_u_generic(int16_t* outputVector, co r = max_val; else if(r < min_val) r = min_val; - *outputVectorPtr++ = (int16_t)(r); + *outputVectorPtr++ = (int16_t)rintf(r); } } #endif /* LV_HAVE_GENERIC */ |