134 const float* aVector,
135 const double* bVector,
136 unsigned int num_points)
138 unsigned int number = 0;
139 const unsigned int eighth_points = num_points / 8;
141 double* cPtr = cVector;
142 const float* aPtr = aVector;
143 const double* bPtr = bVector;
147 __m256d aDbl1, aDbl2, bVal1, bVal2, cVal1, cVal2;
148 for (; number < eighth_points; number++) {
150 aVal = _mm256_loadu_ps(aPtr);
151 bVal1 = _mm256_loadu_pd(bPtr);
152 bVal2 = _mm256_loadu_pd(bPtr + 4);
154 aVal1 = _mm256_extractf128_ps(aVal, 0);
155 aVal2 = _mm256_extractf128_ps(aVal, 1);
157 aDbl1 = _mm256_cvtps_pd(aVal1);
158 aDbl2 = _mm256_cvtps_pd(aVal2);
160 cVal1 = _mm256_add_pd(aDbl1, bVal1);
161 cVal2 = _mm256_add_pd(aDbl2, bVal2);
163 _mm256_storeu_pd(cPtr,
165 _mm256_storeu_pd(cPtr + 4,
173 number = eighth_points * 8;
174 for (; number < num_points; number++) {
175 *cPtr++ = ((double)(*aPtr++)) + (*bPtr++);
187 const float* aVector,
188 const double* bVector,
189 unsigned int num_points)
191 unsigned int number = 0;
192 const unsigned int eighth_points = num_points / 8;
194 double* cPtr = cVector;
195 const float* aPtr = aVector;
196 const double* bPtr = bVector;
200 __m256d aDbl1, aDbl2, bVal1, bVal2, cVal1, cVal2;
201 for (; number < eighth_points; number++) {
203 aVal = _mm256_load_ps(aPtr);
204 bVal1 = _mm256_load_pd(bPtr);
205 bVal2 = _mm256_load_pd(bPtr + 4);
207 aVal1 = _mm256_extractf128_ps(aVal, 0);
208 aVal2 = _mm256_extractf128_ps(aVal, 1);
210 aDbl1 = _mm256_cvtps_pd(aVal1);
211 aDbl2 = _mm256_cvtps_pd(aVal2);
213 cVal1 = _mm256_add_pd(aDbl1, bVal1);
214 cVal2 = _mm256_add_pd(aDbl2, bVal2);
216 _mm256_store_pd(cPtr, cVal1);
217 _mm256_store_pd(cPtr + 4,
225 number = eighth_points * 8;
226 for (; number < num_points; number++) {
227 *cPtr++ = ((double)(*aPtr++)) + (*bPtr++);