[PATCH] tools: add support for decoding ARM attributes

[Saleem Abdulrasool]

  I didn't go that route since the display strings are not something that will be used by clang/LLVM.  So, I didnt quite see the point with bloating Support with the string data.  Im afraid that we will force clang to pay a non-trivial amount of overhead by forcing it to load the data on each invocation if we move the strings into support.

  Secondly, if the string descriptions change, I dont see why that needs to cause changes to anything other than the tool displaying the string (llvm-readobj).

  Maybe Im trying to over-optimise here, but, I dont see a concrete benefit for the movement is all.

http://llvm-reviews.chandlerc.com/D2576