[PATCH] D13994: Optimization for pow(x, n) where n is some constant

[Mandeep Singh Grang via llvm-commits]

mgrang added a comment.

Thanks Chad and apologies to everyone for pinging too early!

I guess doing this transformation for exponents greater than 32 would not really be helpful since the no. of fmul(s) would increase.
So I would like to limit this only to exponents <= 32 (+ve and -ve).

I would then change the command line option to bool (default value true), so that the user can turn off the transformation if he wants to.

Thoughts?


http://reviews.llvm.org/D13994