2.13rc windows x86 编译不带avx指令集的paddlelite失败

[hiweed]

• 版本、环境信息：
     1）Paddle Lite v2.13rc
     2）目标 x86 静态库
     3）编译选项 lite\tools\build_windows.bat with_extra without_python build_x86 with_static_mkl with_dynamic_crt use_vs2017
• 问题描述：附有without_avx的话 编译过程报错，启用avx指令集的编译版本则一切正常

报错如下

 
       “D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\publish_inference.vcxproj”(默认目标) (1) ->
       “D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\publish_inference_x86_cxx_lib.vcxproj”(默认目标) (2) ->
       “D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\api\bundle_full_api.vcxproj”(默认目标) (3) ->
       “D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\api\paddle_api_full.vcxproj”(默认目标) (4) ->
       “D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj”(默认目标) (24) ->
       (ClCompile 目标) ->
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(89): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(89): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(90): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(90): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(91): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(91): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(92): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(92): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(95): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(95): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(96): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(96): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(97): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(97): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(98): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_
       math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(98): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(107): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86
       _math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(107): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(108): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86
       _math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(108): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(109): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86
       _math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(109): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(110): error C2676: 二进制“[”:“__m128”不定义该运算符或到预定义运算符可接收的类型的转换 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86
       _math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\activation.cc(110): error C2088: “[”: 对于 union 非法 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(202): error C3861: “_mm_loadu_epi8”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(203): error C3861: “_mm_loadu_epi8”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(204): error C3861: “_mm_loadu_epi8”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(205): error C3861: “_mm_loadu_epi8”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(213): error C3861: “mm_cvtepi32_ps”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(214): error C3861: “mm_cvtepi32_ps”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(215): error C3861: “mm_cvtepi32_ps”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]
         D:\Work\cpp_libs_src\Paddle-Lite\lite\backends\x86\math\calib.cc(225): error C3861: “_mm_loadu_epi8”: 找不到标识符 [D:\Work\cpp_libs_src\Paddle-Lite\build.lite.x86\lite\backends\x86\x86_math.vcxproj]

[zhupengyang]

部分x86算子必须依赖avx，不能带without_avx。

x86上推理更建议使用PaddleInference