142         const __m128 vi2x2 = _mm_loadu_ps(i2); i2 += 4;  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()  local
149 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk00c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
162 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk20c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
170 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk00c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
183 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk20c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
191 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk01c0, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(2, 2, 2, 2))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
204 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk21c0, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(2, 2, 2, 2))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
212 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk01c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(3, 3, 3, 3))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
225 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk21c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(3, 3, 3, 3))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
280 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk02c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
293 …vo0x0 = _mm_add_ps(vo0x0, _mm_mul_ps(vk22c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
301 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk02c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
314 …vo0x0 = _mm_add_ps(vo0x0, _mm_mul_ps(vk22c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
396         __m128 vi2x2 = _mm_setzero_ps();  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()  local
403           vi2x2 = _mm_loadu_ps(i2 + 4);  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
411 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk00c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
424 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk20c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
432 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk00c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
445 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk20c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
454 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk01c0, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(2, 2, 2, 2))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
471 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk21c0, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(2, 2, 2, 2))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
481 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk01c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(3, 3, 3, 3))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
498 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk21c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(3, 3, 3, 3))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
556 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk02c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
565 …vo0x0 = _mm_add_ps(vo0x0, _mm_mul_ps(vk22c1, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(0, 0, 0, 0))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
571 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk02c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
580 …vo0x0 = _mm_add_ps(vo0x0, _mm_mul_ps(vk22c2, _mm_shuffle_ps(vi2x2, vi2x2, _MM_SHUFFLE(1, 1, 1, 1))…  in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()