naive kernel_conf creation

oneflow/ir/oneflow-jit/JIT.cpp

@@ -254,6 +254,13 @@ llvm::Optional<TensorType> JitImporter::GetMlirTensorTypeFromBlobDesc(const Blob
   }
 }
 
+ParallelContext GetSingleDeviceParallelContext() {
+  ParallelContext parallel_ctx;
+  parallel_ctx.set_parallel_id(0);
+  parallel_ctx.set_parallel_num(1);
+  return parallel_ctx;
+}
+
 void JitImporter::CreateOperandMapping(const ::oneflow::OperatorConf& op_conf,
                                        const std::shared_ptr<const ParallelDesc> parallel_desc,
                                        const std::shared_ptr<const ArgTuple>& input_arg_tuple,
@@ -293,11 +300,16 @@ void JitImporter::CreateOperandMapping(const ::oneflow::OperatorConf& op_conf,
     return lbi2logical_blob_desc_.at(bn).get();
   };
   CHECK_JUST(op->InferLogicalOutBlobDescs(GetLogicalBlobDesc4BnInOp, *parallel_desc));
+  KernelConf kernel_conf;
+  static ParallelContext parallel_ctx = GetSingleDeviceParallelContext();
   for (auto& kv : lbi2logical_blob_desc_) {
     CHECK(
         result_type_mapping_.emplace(kv.first, GetMlirTensorTypeFromBlobDesc(*kv.second).getValue())
             .second);
   }
+  op->GenKernelConf(GetLogicalBlobDesc4BnInOp, &parallel_ctx, &kernel_conf);
+  llvm::errs() << "kernel_conf: \n";
+  llvm::errs() << kernel_conf.DebugString() << "\n";
 }
 
 llvm::Optional<mlir::Value> JitImporter::GetResultByBnAndIndex(const std::string& bn,