Hi Kaffe guys, I want to write a build recipe for kaffe 1.1.8 for OpenEmbedded and do not really know when I need the --enable-xscale configure option (for ARM JIT).
The ARM FAQ says: One more note, current configure option '--enable-xscale' changes the behavior not only for instruction set but also for exception handler. The later is done because of xscale does not have FPU and number of registers saved on the stack is different from with FPU version. I have shown that to Koen Koi one of the OE developers and he said: 'xscale' usually means 'armv5t + dsp extensions' (the 'e' in armv5te) coupled with an absurdly long pipeline. The arm FPU (fpa) usually is on CP0 and CP1 and iwmmxt on CP1 and CP2, so the instructions overlap and you binary goes kaboom. With (soft)vfp and eabi they don't conflict anymore. So I'm not sure what that faq is getting at (armv4 vs armv5te or fpa vs vfp or armv5te vs iwmmxt), but it suggests that --enable-xscale is valid for every xscale arch (pxa2xx, ipx4xx, iopxxx). Would you please add clarifying documentation to the ARM FAQ so that it is more easy to find out whether --enable-xscale is needed or not? Btw: I recently tried compiling kaffe against in a EABI-only environment. I know this is not supported. However the build simply fails (some symbol for cache flushin is not found). It would be better if there is a #ifdef __ARM_EABI__ #error "ARM EABI is not supported yet." #endif File a bug for that? Regards Robert
signature.asc
Description: OpenPGP digital signature
_______________________________________________ kaffe mailing list [email protected] http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
