On Monday 13 July 2009 03:36:39 Stefan Roese wrote: > On Sunday 12 July 2009 12:29:32 Wolfgang Denk wrote: > > > > Any idea why this still happens *with* libgcc patches? Any idea how > > > > to fix this? Add __umoddi3 and __udivdi3 to libgcc patch, too? > > > > > > I have to admit that I'm not sure why this is the case. But I suggest > > > that you take a look at Simon's patch sent to the list a few days ago: > > > > > > [PATCH 5/8]: Use do_div from div64.h for vsprintf > > > > > > This should fix this issue. > > > > It will hush up the current errors, but that's actually a bad thing > > here - the errors are an indication that Jean-Christophe's patch > > might not be working as it is supposed to. > > From my point of view those two patches address different issues. I just > noticed that Simon's patch solved some of the current ARM toolchain > problems as well. That's why I referred to it. > > So I think that regardless of this ARM libgcc discussion, we should apply > Simons patch "[PATCH 5/8]: Use do_div from div64.h for vsprintf" as it uses > the common do_div() implementation instead of a local version.
correct. doing 64bit math is expensive and should be avoided. this is another reason the linux kernel includes libgcc -- they dont import many of the 64bit math functions because it's so expensive and 99% of the time, it's wrong. when the source uses it, they get link failures right away. -mike
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
