Hi All
Can anyone explain the following?
1) I have a reference project that builds with CW7.2, CW10.1 and GCC
2) Each of the compiler targets are set up for maximum optimisation for size (in particular CW7.2 has register passing configured)
3) The code size (that is, the size of the code burned into FLASH) is:
CW 7.2 = 82k
GCC = 88k
CW10.1 = 113k
I would expect that CW 7.2 be a bit better than GCC but I wonder why CW10.1 is generating 38% more code (28% more that GCC). With decreasing optimisation level the code size goes up, so it is not that it is not trying...
Regards
Mark
Hi All
Thanks to Freescale support I compared a few other settings and managed quite an imporvement by setting the following processor options:
- Register coloring
- Instruction scheduling
- Peephole
The CW10.1 code size is now 84k, which is very close to the CW7.2 code size.
Regards
Mark