|
|
|
@@ -11,13 +11,14 @@ |
|
|
|
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR |
|
|
|
// CONDITIONS OF ANY KIND, either express or implied. See the License for the |
|
|
|
// specific language governing permissions and limitations under the License. |
|
|
|
|
|
|
|
#include "bias_x86.h" |
|
|
|
|
|
|
|
#include <emmintrin.h> |
|
|
|
#if __AVX__ |
|
|
|
#include <immintrin.h> |
|
|
|
#endif // __AVX__ |
|
|
|
|
|
|
|
#include "bias_x86.h" |
|
|
|
|
|
|
|
namespace ncnn { |
|
|
|
|
|
|
|
int Bias_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const |
|
|
|
@@ -36,7 +37,6 @@ int Bias_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const |
|
|
|
float bias = bias_ptr[q]; |
|
|
|
|
|
|
|
int i = 0; |
|
|
|
__m128 _bias128; |
|
|
|
#if __AVX__ |
|
|
|
{ |
|
|
|
__m256 _bias256 = _mm256_set1_ps(bias); |
|
|
|
@@ -48,16 +48,14 @@ int Bias_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const |
|
|
|
|
|
|
|
ptr += 8; |
|
|
|
} |
|
|
|
_bias128 = _mm256_castps256_ps128(_bias256); |
|
|
|
} |
|
|
|
#else |
|
|
|
_bias128 = _mm_set1_ps(bias); |
|
|
|
#endif // __AVX__ |
|
|
|
{ |
|
|
|
__m128 _bias = _mm_set1_ps(bias); |
|
|
|
for (; i + 3 < size; i += 4) |
|
|
|
{ |
|
|
|
__m128 _p = _mm_loadu_ps(ptr); |
|
|
|
__m128 _outp = _mm_add_ps(_p, _bias128); |
|
|
|
__m128 _outp = _mm_add_ps(_p, _bias); |
|
|
|
_mm_storeu_ps(ptr, _outp); |
|
|
|
|
|
|
|
ptr += 4; |
|
|
|
|