------- Comment #15 from bernds at gcc dot gnu dot org 2010-04-27 09:47 ------- Thanks. Could you attach those object files (ignoring ifcvt.o since it obviously changes due to the source change)?
Even better would be if you could produce assembly output by finding the command that produced the outputs and adding -S etc. Please also add "-v" to the compiler so that we can see the exact options that are passed to cc1 (as I mentioned, I'm unable to reproduce codegen differences, and my set of testcases includes .i from recent gcc sources, so I'm wondering if there are any machine-specific options I need to pass). -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43858
