Wolfgang Denk wrote: > Dear Scott Wood, > > In message <4a550b66.1090...@freescale.com> you wrote: >> Note that the overhead of 64-bit printf (assuming the 64-bit >> divide/remainder functions in libgcc aren't pulled in for anything else) >> is 2524 bytes on powerpc (using GCC 4.3.2). That's less than NFS, >> which gets turned on by default (as was brought up recently). :-) > > But compare the functionality...
I use 64-bit prints from time to time. I've never used NFS in u-boot. More relevantly, it's obvious if NFS is missing -- it says "Unknown command". It's not always obvious whether (or why) the printf output you're looking at has been corrupted by an incomplete implementation. >> What if we were to invert the option (CONFIG_SYS_NO_64BIT_VSPRINTF) so >> that it would only be disabled on boards at the intersection of tight >> space constraints and a board maintainer who's pretty sure this board >> doesn't need it? There could also be some warning from printf() if %ll >> is used when not supported, and/or it could still check for %ll and pop >> a long long from the varargs but discard the high half. > > And by default we would add CONFIG_SYS_NO_64BIT_VSPRINTF to all board > config files? No, that's the point -- it would require the board maintainer to explicitly say "this board doesn't need this". By default we would provide a correct printf. -Scott _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
