Home
last modified time | relevance | path

Searched refs:_sum0_3 (Results 1 – 5 of 5) sorted by relevance

/dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/x86/
H A Dconvolution_sgemm.h824 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
853 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
854 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
861 _sum0_3 = _mm_fmadd_ps(_va, _vb0, _sum0_3); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
868 _mm_storeu_ps(output_sum0_3, _sum0_3); in conv_im2col_sgemm_sse()
1392 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
1421 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
1422 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
1429 _sum0_3 = _mm_add_ps(_sum0_3, _mm_mul_ps(_va, _vb0)); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
1436 _mm_storeu_ps(sum0_3_tmp, _sum0_3); in conv_im2col_sgemm_sse()
/dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/x86/
H A Dconvolution_sgemm.h824 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
853 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
854 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
861 _sum0_3 = _mm_fmadd_ps(_va, _vb0, _sum0_3); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
868 _mm_storeu_ps(output_sum0_3, _sum0_3); in conv_im2col_sgemm_sse()
1392 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
1421 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
1422 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
1429 _sum0_3 = _mm_add_ps(_sum0_3, _mm_mul_ps(_va, _vb0)); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
1436 _mm_storeu_ps(sum0_3_tmp, _sum0_3); in conv_im2col_sgemm_sse()
/dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/x86/
H A Dconvolution_sgemm.h824 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
853 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
854 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
861 _sum0_3 = _mm_fmadd_ps(_va, _vb0, _sum0_3); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
868 _mm_storeu_ps(output_sum0_3, _sum0_3); in conv_im2col_sgemm_sse()
1392 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
1421 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
1422 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
1429 _sum0_3 = _mm_add_ps(_sum0_3, _mm_mul_ps(_va, _vb0)); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
1436 _mm_storeu_ps(sum0_3_tmp, _sum0_3); in conv_im2col_sgemm_sse()
/dports/misc/ncnn/ncnn-20211208/src/layer/x86/
H A Dconvolution_sgemm.h824 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
853 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
854 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
861 _sum0_3 = _mm_comp_fmadd_ps(_va, _vb0, _sum0_3); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
868 _mm_storeu_ps(output_sum0_3, _sum0_3); in conv_im2col_sgemm_sse()
1392 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
1421 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
1422 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
1429 _sum0_3 = _mm_add_ps(_sum0_3, _mm_mul_ps(_va, _vb0)); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
1436 _mm_storeu_ps(sum0_3_tmp, _sum0_3); in conv_im2col_sgemm_sse()
/dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/x86/
H A Dconvolution_sgemm.h824 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
853 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
854 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
861 _sum0_3 = _mm_fmadd_ps(_va, _vb0, _sum0_3); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
868 _mm_storeu_ps(output_sum0_3, _sum0_3); in conv_im2col_sgemm_sse()
1392 __m128 _sum0_3 = _mm_loadu_ps(biasptr); in conv_im2col_sgemm_sse() local
1421 _sum0_3 = _mm_add_ps(_sum0_3, _sum0); in conv_im2col_sgemm_sse()
1422 _sum0_3 = _mm_add_ps(_sum0_3, _sum2); in conv_im2col_sgemm_sse()
1429 _sum0_3 = _mm_add_ps(_sum0_3, _mm_mul_ps(_va, _vb0)); // sum0 += (k00-k30) * a00 in conv_im2col_sgemm_sse()
1436 _mm_storeu_ps(sum0_3_tmp, _sum0_3); in conv_im2col_sgemm_sse()