c++ - Does using -parallel or (and) -vec will make our program run faster? -
i'm using linux & intel compiler (c/c++).
my code not use cilkplus or openmp. read following article : auto-parallelization overview
and i'm confused:
- if i'm not using openmp (or cilkplus) , compile program "-parallel" (auto-parallelizer) , "-vec" (auto-vectorization), program run faster (because compiler break serial code parallel code ?)
- if (it run faster), why not default option ? (why not use flags ) ?
thanks
intel uses such techniques loop unwinding, task decomposition & etc parallel code. these techniques make program run faster problem may occurred when program faces race condition, deadlock, tasks done priority , .... can not used default.
Comments
Post a Comment