[tuples, patch 7/7] New test cases for modification analysis and indirect inlining

[Paolo Carlini]

Paolo Carlini wrote:
> ....Eventually, I would not expect that, I would expect that when the 
> call is fully-inlined the two cases should perform essentially the 
> same way. Did you notice this? Is there anything I can do to help?
For the record, in the meanwhile I spent a bit of time analyzing this 
issue and I'm happy to report  that Martin' patch works absolutely fine, 
there is nothing to improve, as far as the libstdc++ testcase is 
concerned. The reason for the remaining performance gap is that the 
outer function call cannot be inlined in the slow case, simply because 
the entire template class instantiation to which it belongs 
(basic_ostream<char>) is (intentionally) declared extern inline.

Sorry for the false alarm and thanks again for the excellent work,
Paolo.