Moving floating point through integer registers

[Canqun Yang]

Hi, all

GCC should have implemented the feature of moving 
floating point through integer registers according to 
the comments listed in 
http://gcc.gnu.org/projects/optimize.html#fpmove.

I'm going to test its benefit on IA64 platform for 
floating point benchmarks. But I wonder whether this 
feature has been implemented on IA64. If not, can 
anyone give some advice about how to implement it?
 

Canqun Yang.
Creative Compiler Research Group.
National University of Defense Technology, China.