phenix.reduce doesn't work well in my hands. I used "automatically add H's" in the phenix GUI instead.
PS: phenix.reduce reads my file, but it output a pdb file with the added H's with "0" in the second column (and "new" in the last column)