add implement of resnet_basic_block op for XPU2, test=kunlun (#44143)

paddle/fluid/operators/fused/resnet_basic_block_op.cc

@@ -258,7 +258,8 @@ class ResNetBasicBlockOp : public framework::OperatorWithKernel {
 class ResNetBasicBlockOpMaker : public framework::OpProtoAndCheckerMaker {
  public:
   void Make() {
-    //  has_shortcut = True:         X           else:       X
+    //  has_shortcut = True:       else:
+    //          X                         X
     //        /                         /
     //      |       |                 |       |
     //    CONV1     |               CONV1     |

paddle/fluid/platform/device/xpu/xpu2_op_list.h

@@ -505,6 +505,14 @@ XPUOpMap& get_kl2_ops() {
        XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
       {"sequence_conv_grad",
        XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
+
+      // Fused op
+      {"resnet_basic_block_grad",
+       XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
+                     pOpKernelType(vartype::FP16, XPUPlace())})},
+      {"resnet_basic_block",
+       XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
+                     pOpKernelType(vartype::FP16, XPUPlace())})},
   };
 
   return s_xpu2_kernels;