fix range op crash in dygraph xpu place (#30469)

Author: taixiurong

paddle/fluid/operators/range_op_xpu.cc

@@ -0,0 +1,69 @@
+/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License. */
+
+#ifdef PADDLE_WITH_XPU
+#include "paddle/fluid/operators/range_op.h"
+#include "paddle/fluid/framework/op_registry.h"
+
+namespace paddle {
+namespace operators {
+
+template <typename T>
+class XPURangeKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& context) const override {
+    auto* start_t = context.Input<framework::Tensor>("Start");
+    auto* end_t = context.Input<framework::Tensor>("End");
+    auto* step_t = context.Input<framework::Tensor>("Step");
+    auto* out = context.Output<framework::Tensor>("Out");
+
+    framework::Tensor n;
+    framework::TensorCopy(*start_t, platform::CPUPlace(), &n);
+    T start = n.data<T>()[0];
+    framework::TensorCopy(*end_t, platform::CPUPlace(), &n);
+    T end = n.data<T>()[0];
+    framework::TensorCopy(*step_t, platform::CPUPlace(), &n);
+    T step = n.data<T>()[0];
+
+    int64_t size = 0;
+    GetSize(start, end, step, &size);
+    out->Resize(framework::make_ddim({size}));
+
+    T* out_data = out->mutable_data<T>(context.GetPlace());
+
+    framework::Tensor out_cpu;
+    T* out_cpu_data_ptr =
+        out_cpu.mutable_data<T>(platform::CPUPlace(), out->numel() * sizeof(T));
+    T value = start;
+    for (int64_t i = 0; i < size; ++i) {
+      out_cpu_data_ptr[i] = value;
+      value += step;
+    }
+    int ret = xpu_memcpy(out_data, out_cpu_data_ptr, out->numel() * sizeof(T),
+                         XPUMemcpyKind::XPU_HOST_TO_DEVICE);
+    PADDLE_ENFORCE_EQ(ret, XPU_SUCCESS,
+                      platform::errors::External("XPU xpu_memcpy return wrong "
+                                                 "value[%d %s]",
+                                                 ret, XPUAPIErrorMsg[ret]));
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OP_XPU_KERNEL(range, ops::XPURangeKernel<int>,
+                       ops::XPURangeKernel<int64_t>, ops::XPURangeKernel<float>,
+                       ops::XPURangeKernel<double>);
+#endif  // PADDLE_WITH_XPU

python/paddle/fluid/tests/unittests/xpu/test_range_xpu.py

@@ -0,0 +1,76 @@
+#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+
+import unittest
+import paddle
+import numpy as np
+import sys
+sys.path.append("..")
+from op_test_xpu import XPUOpTest
+
+paddle.enable_static()
+
+
+class TestRangeOp(XPUOpTest):
+    def setUp(self):
+        self.op_type = "range"
+        self.init_config()
+        self.inputs = {
+            'Start': np.array([self.case[0]]).astype(self.dtype),
+            'End': np.array([self.case[1]]).astype(self.dtype),
+            'Step': np.array([self.case[2]]).astype(self.dtype)
+        }
+
+        self.outputs = {
+            'Out': np.arange(self.case[0], self.case[1],
+                             self.case[2]).astype(self.dtype)
+        }
+
+    def init_config(self):
+        self.dtype = np.float32
+        self.case = (0, 1, 0.2)
+
+    def test_check_output(self):
+        place = paddle.XPUPlace(0)
+        self.check_output_with_place(place, check_dygraph=False)
+
+
+class TestFloatRangeOpCase0(TestRangeOp):
+    def init_config(self):
+        self.dtype = np.float32
+        self.case = (0, 5, 1)
+
+
+class TestInt32RangeOpCase0(TestRangeOp):
+    def init_config(self):
+        self.dtype = np.int32
+        self.case = (0, 5, 2)
+
+
+class TestInt32RangeOpCase1(TestRangeOp):
+    def init_config(self):
+        self.dtype = np.int32
+        self.case = (10, 1, -2)
+
+
+class TestInt32RangeOpCase2(TestRangeOp):
+    def init_config(self):
+        self.dtype = np.int32
+        self.case = (-1, -10, -2)
+
+
+if __name__ == "__main__":
+    unittest.main()