[v1.7.x] ElementWiseSum fix for oneDNN (#18777)

* Fix ElementwiseSum for DNNL

* Fix sanity and replace push_back with emplace_back

* Change order of the data format conditions

* Add NOLINT to avoid readability error

* Add test for oneDNN ElemwiseSum

Co-authored-by: Bart Gawrych <gawrych.bartlomiej@intel.com>

Author: bgawrych

src/operator/tensor/elemwise_sum.cc

@@ -113,7 +113,14 @@ void ElementWiseSumComputeExCPU(const nnvm::NodeAttrs& attrs,
   CHECK_EQ(outputs.size(), 1U);
   CHECK_EQ(req.size(), 1U);
   if (req[0] == kNullOp) return;
-  if (common::ContainsOnlyStorage(inputs, kRowSparseStorage) ||
+#if MXNET_USE_MKLDNN == 1
+  if (IsMKLDNNData(inputs)) {
+    MKLDNNRun(MKLDNNSumForward, attrs, ctx, inputs, req, outputs);
+  } else if (common::ContainsOnlyStorage(inputs, kDefaultStorage)) {
+    FallBackCompute(ElementWiseSumCompute<cpu>, attrs, ctx, inputs, req, outputs);
+  }
+#endif
+  else if (common::ContainsOnlyStorage(inputs, kRowSparseStorage) || // NOLINT(*)
       (inputs.size() == 3U && inputs[0].storage_type() == kDefaultStorage &&
        inputs[1].storage_type() == kCSRStorage && inputs[2].storage_type() == kDefaultStorage) ||
       (inputs.size() > 4U && common::ContainsStorageType(inputs, kDefaultStorage) &&
@@ -123,12 +130,6 @@ void ElementWiseSumComputeExCPU(const nnvm::NodeAttrs& attrs,
         ResourceRequest(ResourceRequest::kTempSpace));
     NDArray out_nd = outputs[0];
     mxnet::ndarray::ElementwiseSum<cpu>(s, rsc, inputs, &out_nd);
-#if MXNET_USE_MKLDNN == 1
-  } else if (IsMKLDNNData(inputs)) {
-    MKLDNNRun(MKLDNNSumForward, attrs, ctx, inputs, req, outputs);
-  } else if (common::ContainsOnlyStorage(inputs, kDefaultStorage)) {
-    FallBackCompute(ElementWiseSumCompute<cpu>, attrs, ctx, inputs, req, outputs);
-#endif
   } else {
     LogUnimplementedOp(attrs, ctx, inputs, req, outputs);
   }

tests/python/mkl/test_mkldnn.py

@@ -254,6 +254,28 @@ def test_flatten_slice_after_conv():
     print(p[0])
 
 
+def test_mkldnn_sum_with_mkldnn_layout():
+
+    x_shape = (32, 3, 224, 224)
+    x_npy = np.ones(x_shape)
+    w_shape = (32, 3, 3, 3)
+    w_npy = np.ones(w_shape)
+
+    x = mx.sym.Variable("x")
+    w = mx.sym.Variable("w")
+    z = mx.symbol.Convolution(data=x, weight=w, num_filter=32, kernel=(3, 3))
+    num_inputs = [2, 3, 4, 5]
+    for i in num_inputs:
+        inputs = []
+        for n in range(i):
+            inputs.append(z)
+        y = mx.sym.add_n(*inputs) # (only MKLDNN data input)
+        exe = y.simple_bind(ctx=mx.cpu(), x=x_shape, w=w_shape)
+        out = exe.forward(is_train=False, x=x_npy, w=np.ones(w_shape))[0]
+        #conv with kernel (3,3) on ones should give result=27
+        single_cov = 27.0
+        assert_almost_equal(out[0].asnumpy()[0, 0, 0], single_cov*i)
+
 def test_mkldnn_sum_inplace_with_cpu_layout():
 
     x_shape = (32, 3, 224, 224)
@@ -263,7 +285,7 @@ def test_mkldnn_sum_inplace_with_cpu_layout():
     x = mx.sym.Variable("x")
     y = mx.sym.Variable("y")
     z = mx.symbol.Convolution(data=x, num_filter=32, kernel=(3, 3))
-    z = mx.sym.add_n(z, y)
+    z = mx.sym.add_n(z, y) # (MKLDNN data, cpu data)
     exe = z.simple_bind(ctx=mx.cpu(), x=x_shape, y=y_shape)
     out = exe.forward(is_train=False, x=x_npy, y=y_npy)[0]
     assert_almost_equal(out[0].asnumpy()[0, 0, 0], 1.0)