{"id":43,"date":"2016-10-04T10:57:27","date_gmt":"2016-10-04T10:57:27","guid":{"rendered":"http:\/\/vordhosbn.com\/?p=43"},"modified":"2016-11-23T15:38:59","modified_gmt":"2016-11-23T15:38:59","slug":"gcc-vs-clang-vs-icc-for-32-bit-register-manipulation","status":"publish","type":"post","link":"https:\/\/vordhosbn.com\/index.php\/blog\/cpp\/gcc-vs-clang-vs-icc-for-32-bit-register-manipulation\/","title":{"rendered":"GCC vs Clang vs ICC for 32-bit register manipulation"},"content":{"rendered":"

Today godbolt<\/a>\u00a0released a new compiler comparison feature and I decided to quickly try to compare C++ vs C style register manipulation and see what kind of optimizations will be applied across the different compilers.<\/p>\n

Source code<\/h1>\n

So I wrote a naive\u00a0C++ class that would represent a 32-bit register.<\/p>\n

Note that I am trying here to test compilers’ optimization on class member manipulation, rather than making the case for embedded C++ (where bits in registers would be volatile, since the order of bit setting is significant most of the time).<\/p>\n

struct Register\r\n{\r\nunsigned int b0:1;\r\nunsigned int b1:1;\r\nunsigned int b2:1;\r\nunsigned int b3:1;\r\nunsigned int b4:1;\r\nunsigned int b5:1;\r\nunsigned int b6:1;\r\nunsigned int b7:1;\r\nunsigned int b8:1;\r\nunsigned int b9:1;\r\nunsigned int b10:1;\r\nunsigned int b11:1;\r\nunsigned int b12:1;\r\nunsigned int b13:1;\r\nunsigned int b14:1;\r\nunsigned int b15:1;\r\nunsigned int b16:1;\r\nunsigned int b17:1;\r\nunsigned int b18:1;\r\nunsigned int b19:1;\r\nunsigned int b20:1;\r\nunsigned int b21:1;\r\nunsigned int b22:1;\r\nunsigned int b23:1;\r\nunsigned int b24:1;\r\nunsigned int b25:1;\r\nunsigned int b26:1;\r\nunsigned int b27:1;\r\nunsigned int b28:1;\r\nunsigned int b29:1;\r\nunsigned int b30:1;\r\nunsigned int b31:1;\r\n\r\ninline void Set(unsigned char bitNumber, bool value)\r\n{\r\n    switch(bitNumber)\r\n    {\r\n        case 0: b0 = value; break;\r\n        case 1: b1 = value; break;\r\n        case 2: b2 = value; break;\r\n        case 3: b3 = value; break;\r\n        case 4: b4 = value; break;\r\n        case 5: b5 = value; break;\r\n        case 6: b6 = value; break;\r\n        case 7: b7 = value; break;\r\n        case 8: b8 = value; break;\r\n        case 9: b9 = value; break;\r\n        case 10: b10 = value; break;\r\n        case 11: b11 = value; break;\r\n        case 12: b12 = value; break;\r\n        case 13: b13 = value; break;\r\n        case 14: b14 = value; break;\r\n        case 15: b15 = value; break;\r\n        case 16: b16 = value; break;\r\n        case 17: b17 = value; break;\r\n        case 18: b18 = value; break;\r\n        case 19: b19 = value; break;\r\n        case 20: b20 = value; break;\r\n        case 21: b21 = value; break;\r\n        case 22: b22 = value; break;\r\n        case 23: b23 = value; break;\r\n        case 24: b24 = value; break;\r\n        case 25: b25 = value; break;\r\n        case 26: b26 = value; break;\r\n        case 27: b27 = value; break;\r\n        case 28: b28 = value; break;\r\n        case 29: b29 = value; break;\r\n        case 30: b30 = value; break;\r\n        case 31: b31 = value; break;\r\n    }\r\n}\r\n};<\/pre>\n
After that you can see I am using it to set the 0, 1, 3, 5, 13 and 30th bits of the register:<\/p>\n
void cppFunction(Register& reg) \r\n{\r\n  reg.Set(0,true);\r\n  reg.Set(1,true);\r\n  reg.Set(3,true);\r\n  reg.Set(5,true);\r\n  reg.Set(13, true);\r\n  reg.Set(30, true);\r\n}<\/pre>\n
the corresponding C\u00a0implementation is:<\/p>\n
void cFunction(int* reg)\r\n{\r\n  *reg |= (1 << 0) | (1 << 1) | (1 << 3) | (1 << 5) | (1 << 13) | (1 << 30);\r\n}<\/pre>\n
I am also calling the above functions from a third one to see if\u00a0the compilers will inline it:<\/p>\n
void Caller(Register& reg1, int *reg2)\r\n{\r\n  cppFunction(reg1);\r\n  cFunction(reg2);\r\n}<\/pre>\n
The compilers<\/h1>\n
I am comparing\u00a0three compilers:<\/p>\n