Is anyone using the ARM launchpad gcc 4.7 compiler: https://launchpad.net/gcc-arm-embedded
There are several options for code generation for the STM32F3/4, particularly for floating point.
There is also an option to do "Whole Program Optimisation". The GCC documentation says it offers smaller faster binaries through a further optimisation pass at link time. Does this work well? Is it reasonably quick? Does it remove dead-code? Any disadvantages?
What compile and link options are people, who have experimented with it, using?