Date: Sat, 12 Apr 2003 17:31:23 -0700 (PDT) From: Marcel Moolenaar <marcel@FreeBSD.org> To: Perforce Change Reviews <perforce@freebsd.org> Subject: PERFORCE change 28850 for review Message-ID: <200304130031.h3D0VNSe079063@repoman.freebsd.org>
next in thread | raw e-mail | index | archive | help
http://perforce.freebsd.org/chv.cgi?CH=28850 Change 28850 by marcel@marcel_nfs on 2003/04/12 17:31:00 AARRGGHH.. GCC checks the code for calls to "special" functions, such as setjmp(), longjmp() and alloca(). It also checks for savectx() and marks the function as having a call that can return twice. I don't particularly have a problem with that, but it also means that the function has a large prologue and epilogue in which most of the context is saved and restore. Since cpu_switch() is in C and uses savectx() and restorectx() to handle the register contexts, we're doing most of it twice. That sucks :-( For some reason GCC continues to make assumptions about functions even though we compile with -ffreestanding. Stop GCC's interference by renaming savectx to savectx__ with a preprocessor define. We may want to make the define dependent on the compiler, but I didn't bother doing that now... Affected files ... .. //depot/projects/ia64_epc/sys/ia64/ia64/context.s#10 edit .. //depot/projects/ia64_epc/sys/ia64/include/pcb.h#6 edit Differences ... ==== //depot/projects/ia64_epc/sys/ia64/ia64/context.s#10 (text+ko) ==== @@ -163,11 +163,11 @@ END(restorectx) /* - * void savectx(struct pcb *) + * void savectx__(struct pcb *) * void swapctx(struct pcb *old, struct pcb *new) */ -ENTRY(savectx,1) +ENTRY(savectx__,1) { .mmi alloc r16=ar.pfs,1,1,0,0 ;; ==== //depot/projects/ia64_epc/sys/ia64/include/pcb.h#6 (text+ko) ==== @@ -49,6 +49,7 @@ }; #ifdef _KERNEL +#define savectx savectx__ void restorectx(struct pcb *) __dead2; int savectx(struct pcb *); void swapctx(struct pcb *old, struct pcb *new);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200304130031.h3D0VNSe079063>