diff --git a/cmake/external/xpu.cmake b/cmake/external/xpu.cmake
index 45a76fdc1f1a2aab66e7f4972eecbbec03af941a..cfbe68eecbaca55c5a288aae2c985bbc33d37be2 100644
--- a/cmake/external/xpu.cmake
+++ b/cmake/external/xpu.cmake
@@ -36,7 +36,7 @@ ENDIF()
 
 if(NOT DEFINED XPU_BASE_URL)
   SET(XPU_BASE_URL_WITHOUT_DATE "https://baidu-kunlun-product.cdn.bcebos.com/KL-SDK/klsdk-dev")
-  SET(XPU_BASE_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220228")
+  SET(XPU_BASE_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220307")
 else()
   SET(XPU_BASE_URL "${XPU_BASE_URL}")
 endif()
diff --git a/paddle/fluid/operators/tril_triu_op_xpu.cc b/paddle/fluid/operators/tril_triu_op_xpu.cc
new file mode 100644
index 0000000000000000000000000000000000000000..e36cbcf228cfbf30c8fcd5562ac40f38a5467cdb
--- /dev/null
+++ b/paddle/fluid/operators/tril_triu_op_xpu.cc
@@ -0,0 +1,53 @@
+/* Copyright (c) 2022 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/tril_triu_op.h"
+#include "paddle/fluid/platform/device/device_wrapper.h"
+
+namespace paddle {
+namespace operators {
+
+using Tensor = framework::Tensor;
+
+template <typename DeviceContext, typename T>
+class TrilTriuXPUKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& context) const override {
+    const auto* x = context.Input<framework::Tensor>("X");
+    const auto* x_data = x->data<T>();
+    auto* out = context.Output<framework::Tensor>("Out");
+    auto* out_data = out->mutable_data<T>(context.GetPlace());
+
+    const int diagonal = context.Attr<int>("diagonal");
+    const bool lower = context.Attr<bool>("lower");
+    auto xshape = phi::vectorize<int>(x->dims());
+    auto& dev_ctx = context.template device_context<DeviceContext>();
+    int r = 0;
+    if (lower) {
+      r = xpu::tril(dev_ctx.x_context(), x_data, out_data, xshape, diagonal);
+      PADDLE_ENFORCE_XDNN_SUCCESS(r, "tril_op");
+    } else {
+      r = xpu::triu(dev_ctx.x_context(), x_data, out_data, xshape, diagonal);
+      PADDLE_ENFORCE_XDNN_SUCCESS(r, "triu_op");
+    }
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OP_XPU_KERNEL(
+    tril_triu, ops::TrilTriuXPUKernel<paddle::platform::XPUDeviceContext, int>,
+    ops::TrilTriuXPUKernel<paddle::platform::XPUDeviceContext, float>);
+#endif
diff --git a/paddle/fluid/platform/device/xpu/xpu2_op_list.h b/paddle/fluid/platform/device/xpu/xpu2_op_list.h
index e6b08ed7bc340b5150078fe0deb6a3187fb8e17b..3789ec322ac9952011764dd5230b72eb4b9ada19 100644
--- a/paddle/fluid/platform/device/xpu/xpu2_op_list.h
+++ b/paddle/fluid/platform/device/xpu/xpu2_op_list.h
@@ -323,6 +323,8 @@ XPUOpMap& get_kl2_ops() {
        XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
       {"split", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
                               pOpKernelType(vartype::INT32, XPUPlace())})},
+      {"square_grad", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
+      {"square", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
       {"squeeze2_grad",
        XPUKernelSet({pOpKernelType(vartype::FP64, XPUPlace()),
                      pOpKernelType(vartype::INT64, XPUPlace()),
@@ -349,6 +351,8 @@ XPUOpMap& get_kl2_ops() {
                                   pOpKernelType(vartype::FP16, XPUPlace())})},
       {"tanh", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
                              pOpKernelType(vartype::FP16, XPUPlace())})},
+      {"tril_triu", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
+                                  pOpKernelType(vartype::INT32, XPUPlace())})},
       {"tile", XPUKernelSet({pOpKernelType(vartype::INT32, XPUPlace()),
                              pOpKernelType(vartype::INT64, XPUPlace()),
                              pOpKernelType(vartype::BOOL, XPUPlace()),
diff --git a/python/paddle/fluid/tests/unittests/xpu/test_tril_triu_op_xpu.py b/python/paddle/fluid/tests/unittests/xpu/test_tril_triu_op_xpu.py
new file mode 100644
index 0000000000000000000000000000000000000000..785549abba8f3ad732dd49263532fcf453f0517b
--- /dev/null
+++ b/python/paddle/fluid/tests/unittests/xpu/test_tril_triu_op_xpu.py
@@ -0,0 +1,143 @@
+#   Copyright (c) 2022 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 sys
+sys.path.append("..")
+
+import paddle
+import paddle.fluid.core as core
+import paddle.fluid as fluid
+import paddle.tensor as tensor
+import unittest
+import numpy as np
+from op_test import OpTest
+from op_test_xpu import XPUOpTest
+from paddle.fluid.framework import Program, program_guard
+from xpu.get_test_cover_info import create_test_class, get_xpu_op_support_types, XPUOpTestWrapper
+
+paddle.enable_static()
+
+
+class XPUTestTrilTriuOp(XPUOpTestWrapper):
+    def __init__(self):
+        self.op_name = 'tril_triu'
+        self.use_dynamic_create_class = False
+
+    class TestTrilTriuOp(XPUOpTest):
+        def setUp(self):
+            self.init_dtype()
+            self.initTestCase()
+            self.real_op_type = np.random.choice(['triu', 'tril'])
+            self.real_np_op = getattr(np, self.real_op_type)
+            self.set_xpu()
+            self.op_type = "tril_triu"
+            if self.dtype == np.int32:
+                self.X = np.arange(
+                    1, self.get_Xshape_prod() + 1,
+                    dtype=self.dtype).reshape(self.Xshape)
+            else:
+                self.X = np.random.random(self.Xshape).astype(dtype=self.dtype)
+            self.inputs = {'X': self.X}
+            self.attrs = {
+                'diagonal': self.diagonal,
+                'lower': True if self.real_op_type == 'tril' else False,
+            }
+            self.outputs = {
+                'Out': self.real_np_op(self.X, self.diagonal)
+                if self.diagonal else self.real_np_op(self.X)
+            }
+
+        def init_dtype(self):
+            self.dtype = self.in_type
+
+        def get_Xshape_prod(self):
+            ret = 1
+            for v in self.Xshape:
+                ret *= v
+            return ret
+
+        def set_xpu(self):
+            self.__class__.use_xpu = True
+            self.__class__.no_need_check_grad = True
+            self.__class__.op_type = self.real_op_type
+
+        def test_check_output(self):
+            if paddle.is_compiled_with_xpu():
+                place = paddle.XPUPlace(0)
+                self.check_output_with_place(place)
+
+        def initTestCase(self):
+            self.diagonal = None
+            self.Xshape = (10, 10)
+
+    class TestTrilTriuOp1(TestTrilTriuOp):
+        def initTestCase(self):
+            self.diagonal = -3
+            self.Xshape = (5, 5)
+
+    class TestTrilTriuOp2(TestTrilTriuOp):
+        def initTestCase(self):
+            self.diagonal = 4
+            self.Xshape = (11, 17)
+
+    class TestTrilTriuOp3(TestTrilTriuOp):
+        def initTestCase(self):
+            self.diagonal = 10
+            self.Xshape = (25, 25)
+
+    class TestTrilTriuOp4(TestTrilTriuOp):
+        def initTestCase(self):
+            self.diagonal = -10
+            self.Xshape = (33, 11)
+
+    class TestTrilTriuOp5(TestTrilTriuOp):
+        def initTestCase(self):
+            self.diagonal = 11
+            self.Xshape = (1, 99)
+
+
+class TestTrilTriuOpError(unittest.TestCase):
+    def test_errors1(self):
+        paddle.enable_static()
+        data = fluid.data(shape=(20, 22), dtype='float32', name="data1")
+        op_type = np.random.choice(['triu', 'tril'])
+        errmsg = {
+            "diagonal: TypeError":
+            "diagonal in {} must be a python Int".format(op_type),
+        }
+        expected = list(errmsg.keys())[0]
+        with self.assertRaisesRegex(
+                eval(expected.split(':')[-1]), errmsg[expected]):
+            getattr(tensor, op_type)(x=data, diagonal='2022')
+
+    def test_errors2(self):
+        paddle.enable_static()
+        data = fluid.data(shape=(200, ), dtype='float32', name="data2")
+        op_type = np.random.choice(['triu', 'tril'])
+        errmsg = {
+            "input: ValueError":
+            "x shape in {} must be at least 2-D".format(op_type),
+        }
+        expected = list(errmsg.keys())[0]
+        with self.assertRaisesRegex(
+                eval(expected.split(':')[-1]), errmsg[expected]):
+            getattr(tensor, op_type)(x=data, diagonal=[None])
+
+
+support_types = get_xpu_op_support_types('tril_triu')
+for stype in support_types:
+    create_test_class(globals(), XPUTestTrilTriuOp, stype)
+
+if __name__ == '__main__':
+    unittest.main()