144         const __m128 vi4x2 = _mm_loadu_ps(i4); i4 += 4;  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()  local
163 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk20c1, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
184 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk20c2, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
205 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk21c0, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(2, 2, 2, 2))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
226 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk21c1, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(3, 3, 3, 3))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
294 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk22c1, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
315 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk22c2, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
398         __m128 vi4x2 = _mm_setzero_ps();  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()  local
405           vi4x2 = _mm_loadu_ps(i4 + 4);  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
425 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk20c1, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
446 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk20c2, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
472 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk21c0, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(2, 2, 2, 2))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
499 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk21c1, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(3, 3, 3, 3))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
566 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk22c1, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
581 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk22c2, _mm_shuffle_ps(vi4x2, vi4x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()