Unify the TG_DEBUG_REF, update online docs (#637)

* Fix, Call to function with fewer arguments than declared parameters

* update the online doc

README.md

@@ -31,8 +31,11 @@ Tengine 核心代码由 4 个模块组成：
 
 ![Tengine 架构](doc/architecture.png)
 
+## 快速上手
 
-## 如何使用
+### 在线文档
+
+- [Tengine 用户使用手册](https://tengine-docs.readthedocs.io/zh_CN/latest/)
 
 ### 编译
 

README_EN.md

@@ -36,6 +36,10 @@ The core code of Tengine Lite consists of 4 modules:
 
 ## How to use
 
+### Documentation
+
+- [Tengine User Manual](https://tengine-docs.readthedocs.io/en/latest/)
+
 ### Compile
 
 - [Quick Compilation](doc/compile.md) Simple cross-platform compilation based on cmake.

source/device/cpu/cpu_define.h

@@ -39,7 +39,7 @@
 
 #define TENGINE_DUMP_DIR            "TG_DEBUG_DUMP_DIR"
 #define TENGINE_PRINT_LAYER_COST    "TG_DEBUG_TIME"
-#define TENGINE_FORCE_USE_REF_OP    "TG_DEBUG_SCORE"
+#define TENGINE_FORCE_USE_REF_OP    "TG_DEBUG_REF"
 
 
 typedef struct cpu_option