Blue Swirl wrote: > On 4/16/07, Daniel Jacobowitz <[EMAIL PROTECTED]> wrote: > >On Sun, Apr 15, 2007 at 10:37:08PM +0300, Blue Swirl wrote: > >> On 4/15/07, Paul Brook <[EMAIL PROTECTED]> wrote: > >> > On Sunday 15 April 2007 20:11, Blue Swirl wrote: > >> > > > Probably the linker is making sure the file offset and VMA are the > >same > >> > > > modulo the page size. > >> > > > >> > > But that would be one huge file, as the VMA is near 2TB: > >> > > >> > I said *modulo the pace size* :-) > >> > Lets say ld thinks the page size for your system is 1Mb (nor an > >unreasonable > >> > assumption). The vma of .text is aligned on a 1Mb boundary. In order > >to > >> > allow loading via mmap, the location of .text within the file must > >also be > >> > aligned on a 1Mb boundary. It can't put it at address zero because the > >ELF > >> > headers get in the way, so the first viable location is 1Mb into the > >file. > >> > >> Nice theory (and I missed the modulo arithmetic, sorry), but on > >> Ultrasparc the page sizes available are 8k, 64k, 4M and 256M. > > > >#define ELF_MAXPAGESIZE 0x100000 > > > >BFD and GNU ld think it's 1MB. > > I stand corrected. Is there anything that can be done to reduce this waste?
See ld's -z max-page-size and -z common-page-size. Thiemo
