在Linux x86_64上编译llvm：关于优化标志的建议？

Question

I am compiling llvm v3.0 on a linux x86_64 on an intel double core and I would like to have some feedback on the optimization flags that need to be specified. I am compiling with gcc and I would start with '-O3', but I would like to know if there are any other flags that you recommend - maybe also architecture dependent.

thanks

Answer 1

In case you are going to use llvm on the same machine as you are compiling it, I suggest

-O3 -march=native -mtune=native

To enable all supported features of your cpu eg sse,avx,…