Why does cublas still choose column major format instead of row major for matrix?

Cublas doc said it is because of blas traditional format(like Fortran), but nowadays since most of applications is written in c/c++, it is row major(I think), is it necessary to choose column major as default format which is hard to understand for c/c++ programmers, why not change it to row major format?

once you have a code base, making such a change would break nearly everything out there.

yes, I know what you mean. To be compatible with traditional column major format, programmers should pay a little extra attention to format convertion when they are using row major format. I think it will become more friendly if the cublas lib provide a new api or add a parameter to let programmer choose row major directly, maybe this idea lacks deep thinking, :).