> > I now have a functional patch for EABI on v4 processors. I can push it > > out if you're interested. > > I'm exceedingly interested, largely on behalf of people who are using > balloon2 (and thus strongarm) in current production hardware, and > would like to be able to use eabi. > > I can also test stuff on my netwinder.
Patch is in upstream binutils CVS[1]. You'll need additional gcc changes to pass as --fix-v4bx and ld --fix-v4bx-interworking. SUBTARGET_EXTRA_ASM_SPEC and LINK_SPEC in gcc/config/arm/bpabi.h should do the trick. I recommend configuring the toolchain for ARMv4t, and relying on the assembler/linker to fix things up. Configuring for ARMv4 may cause some hand written assembly in glibc/libgcc to incorrectly disable interworking. I have not measured performance overhead. My guess is that branch prediction on modern cores probably folds the additional branch. There is some scope for having gcc inline the veneers. This will probably improve performance at the expense of code size. Paul [1] http://sourceware.org/ml/binutils/2008-02/msg00196.html -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]