MooreThreads · Zhang-joey · Apr 10, 2026
diff --git a/musa_ext/kernels/array/musa_empty_tensor_list.cc b/musa_ext/kernels/array/musa_empty_tensor_list.cc
@@ -10,7 +10,7 @@ Status TensorShapeFromTensor(const Tensor& t, PartialTensorShape* out) {
     if ((t.dtype() == DT_INT32 && t.scalar<int32_t>()() == -1) ||
         (t.dtype() == DT_INT64 && t.scalar<int64_t>()() == -1)) {
       *out = PartialTensorShape();
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
     return errors::InvalidArgument(
         "The only valid scalar shape tensor is the fully unknown shape "

diff --git a/musa_ext/kernels/array/musa_fill_functor.h b/musa_ext/kernels/array/musa_fill_functor.h
@@ -41,7 +41,7 @@ inline Status MusaFillCall(mTensor* out_mt, T value, OpKernelContext* context) {
     return errors::Internal("mudnn run op error!");
   }
 
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 struct SetZeroFunctor {

diff --git a/musa_ext/kernels/array/musa_fill_op.cc b/musa_ext/kernels/array/musa_fill_op.cc
@@ -46,7 +46,7 @@ struct is_any<T, First, Rest...>
 //     return errors::Internal("mtdnn run op error!");
 //   }
 
-//   return Status::OK();
+//   return ::tsl::OkStatus();
 // }
 
 }  // namespace

diff --git a/musa_ext/kernels/array/musa_tensordot_bias_op.cc b/musa_ext/kernels/array/musa_tensordot_bias_op.cc
@@ -55,7 +55,7 @@ REGISTER_OP("MusaTensorDotBias")
 
       if (!c->RankKnown(a_shape) || !c->RankKnown(b_shape)) {
         c->set_output(0, c->UnknownShape());
-        return Status::OK();
+        return ::tsl::OkStatus();
       }
 
       std::vector<int> axes_a, axes_b;
@@ -92,7 +92,7 @@ REGISTER_OP("MusaTensorDotBias")
       }
 
       c->set_output(0, c->MakeShape(output_dims));
-      return Status::OK();
+      return ::tsl::OkStatus();
     });
 
 // =============================================================================
@@ -198,7 +198,7 @@ Status ComputeTensorDotDims(const TensorShape& a_shape,
     }
   }
 
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 template <typename T>
@@ -345,7 +345,7 @@ class MusaTensorDotBiasOp : public MusaOpKernel {
                               static_cast<int>(status));
     }
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   Status DoTensorDotBias(OpKernelContext* ctx, const Tensor& a, const Tensor& b,
@@ -391,11 +391,11 @@ class MusaTensorDotBiasOp : public MusaOpKernel {
       if (!output->CopyFrom(matmul_temp, output_shape)) {
         return errors::Internal("Failed to reshape matmul result to output");
       }
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     TF_RETURN_IF_ERROR(DoMatMulWithBias(ctx, a_2d, b_2d, bias_2d, &matmul_view));
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   // 准备张量：transpose + reshape 为 2D
@@ -434,7 +434,7 @@ class MusaTensorDotBiasOp : public MusaOpKernel {
       }
     }
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   // 准备 bias 张量：reshape 为 1D 向量 (N 维度)
@@ -448,15 +448,15 @@ class MusaTensorDotBiasOp : public MusaOpKernel {
       if (!output->CopyFrom(bias, bias.shape())) {
         return errors::Internal("Failed to copy bias");
       }
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     // 如果 bias 是标量，需要 broadcast 到向量
     if (bias.NumElements() == 1) {
       TensorShape target_shape({n_dim});
       TF_RETURN_IF_ERROR(ctx->allocate_temp(bias.dtype(), target_shape, output));
       // TODO: 实现标量到向量的 broadcast
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     // 如果 bias 是 2D [M, N]，需要提取 N 维度作为向量
@@ -465,7 +465,7 @@ class MusaTensorDotBiasOp : public MusaOpKernel {
       TensorShape target_shape({n_dim});
       TF_RETURN_IF_ERROR(ctx->allocate_temp(bias.dtype(), target_shape, output));
       // TODO: 实现从 2D 提取 1D 的逻辑
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     // 尝试直接 reshape
@@ -476,7 +476,7 @@ class MusaTensorDotBiasOp : public MusaOpKernel {
           " to ", target_shape.DebugString());
     }
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 };
 

diff --git a/musa_ext/kernels/array/musa_tensordot_op.cc b/musa_ext/kernels/array/musa_tensordot_op.cc
@@ -53,7 +53,7 @@ REGISTER_OP("MusaTensorDot")
 
       if (!c->RankKnown(a_shape) || !c->RankKnown(b_shape)) {
         c->set_output(0, c->UnknownShape());
-        return Status::OK();
+        return ::tsl::OkStatus();
       }
 
       std::vector<int> axes_a, axes_b;
@@ -90,7 +90,7 @@ REGISTER_OP("MusaTensorDot")
       }
 
       c->set_output(0, c->MakeShape(output_dims));
-      return Status::OK();
+      return ::tsl::OkStatus();
     });
 
 // =============================================================================
@@ -196,7 +196,7 @@ Status ComputeTensorDotDims(const TensorShape& a_shape,
     }
   }
 
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 template <typename T>
@@ -320,7 +320,7 @@ class MusaTensorDotOp : public MusaOpKernel {
                               static_cast<int>(status));
     }
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   Status DoTensorDot(OpKernelContext* ctx, const Tensor& a, const Tensor& b,
@@ -361,11 +361,11 @@ class MusaTensorDotOp : public MusaOpKernel {
       if (!output->CopyFrom(matmul_temp, output_shape)) {
         return errors::Internal("Failed to reshape matmul result to output");
       }
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     TF_RETURN_IF_ERROR(DoMatMul(ctx, a_2d, b_2d, &matmul_view));
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   // 准备张量：transpose + reshape 为 2D
@@ -404,7 +404,7 @@ class MusaTensorDotOp : public MusaOpKernel {
       }
     }
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 };
 

diff --git a/musa_ext/kernels/array/musa_tensorinteraction_op.cc b/musa_ext/kernels/array/musa_tensorinteraction_op.cc
@@ -97,6 +97,6 @@ REGISTER_OP("MusaInteract")
       ::tensorflow::shape_inference::DimensionHandle n = c->Dim(input_shape, 1);
 
       c->set_output(0, c->MakeShape({batch, n, n}));
-      return Status::OK();
+      return ::tsl::OkStatus();
     });
 }
diff --git a/musa_ext/kernels/array/musa_tensorlist_fromtensor_op.cc b/musa_ext/kernels/array/musa_tensorlist_fromtensor_op.cc
@@ -18,11 +18,11 @@ Status TensorShapeFromTensorMusa(const Tensor& t, PartialTensorShape* out) {
     if (t.dtype() == DT_INT32) {
       out->Clear();
       out->AddDim(static_cast<int64_t>(t.scalar<int32>()()));
-      return Status::OK();
+      return ::tsl::OkStatus();
     } else if (t.dtype() == DT_INT64) {
       out->Clear();
       out->AddDim(static_cast<int64_t>(t.scalar<int64>()()));
-      return Status::OK();
+      return ::tsl::OkStatus();
     } else {
       return errors::InvalidArgument(
           "element_shape must be int32 or int64, got ",
@@ -43,15 +43,15 @@ Status TensorShapeFromTensorMusa(const Tensor& t, PartialTensorShape* out) {
     for (int i = 0; i < vec.size(); ++i) {
       out->AddDim(static_cast<int64_t>(vec(i)));
     }
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   if (t.dtype() == DT_INT64) {
     auto vec = t.vec<int64>();
     for (int i = 0; i < vec.size(); ++i) {
       out->AddDim(static_cast<int64_t>(vec(i)));
     }
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   return errors::InvalidArgument("element_shape must be int32 or int64, got ",

diff --git a/musa_ext/kernels/array/musa_tensorlist_reserve_op.cc b/musa_ext/kernels/array/musa_tensorlist_reserve_op.cc
@@ -19,7 +19,7 @@ Status TensorShapeFromTensorReserve(const Tensor& t, PartialTensorShape* out) {
     if ((t.dtype() == DT_INT32 && t.scalar<int32_t>()() == -1) ||
         (t.dtype() == DT_INT64 && t.scalar<int64_t>()() == -1)) {
       *out = PartialTensorShape();  // Fully unknown shape
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
     return errors::InvalidArgument(
         "The only valid scalar shape tensor is the fully unknown shape "

diff --git a/musa_ext/kernels/array/musa_tokenmixer_op.cc b/musa_ext/kernels/array/musa_tokenmixer_op.cc
@@ -113,7 +113,7 @@ REGISTER_OP("MusaTokenMixer")
       auto batch_dim = c->Dim(c->input(0), 0);
       c->set_output(0, c->MakeShape({batch_dim, c->MakeDim(num_H),
                                      c->MakeDim(num_T * d_k)}));
-      return Status::OK();
+      return ::tsl::OkStatus();
     });
 
 }  // namespace tensorflow
diff --git a/musa_ext/kernels/array/musa_transpose_functor.h b/musa_ext/kernels/array/musa_transpose_functor.h
@@ -22,7 +22,7 @@ struct TransposeFunctor {
     if (::musa::dnn::Status::SUCCESS != pop.Run(h, out_mt, in_mt)) {
       return errors::Internal("muDNN Permute Run failed!");
     }
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 };
 

diff --git a/musa_ext/kernels/array/musa_where_op.h b/musa_ext/kernels/array/musa_where_op.h
@@ -46,7 +46,7 @@ struct NumTrue {
 
     if (input.size() == 0) {
       *num_true_data = static_cast<TIndex>(0);
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     // Use the new LaunchIsNonZeroCount operator which directly counts
@@ -70,7 +70,7 @@ struct NumTrue {
                               musaGetErrorString(m_err));
     }
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 };
 
@@ -99,7 +99,7 @@ struct Where {
                         typename TTypes<T, NDIM>::ConstTensor input,
                         typename TTypes<TIndex>::Matrix output) {
     if (output.dimension(0) == 0) {
-      return Status::OK();
+      return ::tsl::OkStatus();
     }
 
     musaStream_t stream = GetMusaStreamByCtx(ctx);
@@ -171,7 +171,7 @@ struct Where {
     LaunchPropagateWhereIndicesKernel<NDIM, TIndex>(
         output_rows, strides.data(), selected_indices, output.data(), stream);
 
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 };
 

diff --git a/musa_ext/kernels/math/musa_Maxpool_op.cc b/musa_ext/kernels/math/musa_Maxpool_op.cc
@@ -50,7 +50,7 @@ Status PermuteTensorOnMusa(OpKernelContext* ctx, const Tensor& input,
                             static_cast<int>(status));
   }
 
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 Status ComputeOutputAndPadding2D(int64_t in_h, int64_t in_w, int64_t window_h,
@@ -65,7 +65,7 @@ Status ComputeOutputAndPadding2D(int64_t in_h, int64_t in_w, int64_t window_h,
     *pad_bottom = 0;
     *pad_left = 0;
     *pad_right = 0;
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   if (padding == Padding::SAME) {
@@ -81,7 +81,7 @@ Status ComputeOutputAndPadding2D(int64_t in_h, int64_t in_w, int64_t window_h,
     *pad_bottom = static_cast<int>(pad_h - *pad_top);
     *pad_left = static_cast<int>(pad_w / 2);
     *pad_right = static_cast<int>(pad_w - *pad_left);
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   return errors::InvalidArgument(
@@ -119,7 +119,7 @@ Status RunMusaMaxPool(OpKernelContext* ctx, const Tensor& input, Tensor* output,
                             static_cast<int>(status));
   }
 
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 }  // namespace

diff --git a/musa_ext/kernels/math/musa_add_op.cc b/musa_ext/kernels/math/musa_add_op.cc
@@ -229,7 +229,7 @@ Status ConfigureBroadcastView(const Tensor& tensor,
   // Express TensorFlow-style broadcast as a muDNN tensor view by keeping the
   // output dims and setting broadcasted axes to stride 0.
   if (SameShape(tensor, output_shape) || output_shape.dims() == 0) {
-    return Status::OK();
+    return ::tsl::OkStatus();
   }
 
   const int input_rank = tensor.dims();
@@ -265,7 +265,7 @@ Status ConfigureBroadcastView(const Tensor& tensor,
     return errors::Internal("MUSA Add SetNdInfo failed. Status: ",
                             static_cast<int>(status));
   }
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 }  // namespace

diff --git a/musa_ext/kernels/math/musa_cast_functor.h b/musa_ext/kernels/math/musa_cast_functor.h
@@ -22,7 +22,7 @@ static Status CastFunctor(OpKernelContext* ctx, const mTensor& input_mt,
     return errors::Internal("CastTensor Run failed. Status: ",
                             static_cast<int>(status));
   }
-  return Status::OK();
+  return ::tsl::OkStatus();
 }
 
 }  // namespace musa

diff --git a/musa_ext/kernels/math/musa_clip_op.cc b/musa_ext/kernels/math/musa_clip_op.cc
@@ -91,7 +91,7 @@ REGISTER_OP("MusaClip")
       if (!c->RankKnown(x_shape) || !c->RankKnown(lo_shape) ||
           !c->RankKnown(hi_shape)) {
         c->set_output(0, c->UnknownShape());
-        return Status::OK();
+        return ::tsl::OkStatus();
       }
 
       auto BroadcastTwoShapes =
@@ -136,14 +136,14 @@ REGISTER_OP("MusaClip")
 
         std::reverse(dims.begin(), dims.end());
         *out = c->MakeShape(dims);
-        return Status::OK();
+        return ::tsl::OkStatus();
       };
 
       TF_RETURN_IF_ERROR(BroadcastTwoShapes(x_shape, lo_shape, &x_lo_shape));
       TF_RETURN_IF_ERROR(BroadcastTwoShapes(x_lo_shape, hi_shape, &out_shape));
 
       c->set_output(0, out_shape);
-      return Status::OK();
+      return ::tsl::OkStatus();
     });
 
 }  // namespace tensorflow