PGI does target several AMD GPU and APUs using OpenACC directives (current those base on the Tahiti, Cape Verde, and Spectre architectures). Though you will need to add the directive around the compute intensive loops you wish to accelerate so some rewriting will be necessary. However, directives are treated as comments by compilers which don’t recognize them so the code would still be able to be compiled by other compilers.
I would suggest looking at our OpenACC page (http://www.pgroup.com/resources/accel.htm) and spending a bit of time watching a webinar or two. You can then decide if OpenACC is a good fit for what you wish to accomplish.