When building for powerpc-linux-gnuspe target (e500 based platforms) gcc defaults to producing both compiler and binaries with only single precision FP instructions. To get it build with proper double precision FP, an additional switch should be passed to "configure": "--enable-e500-double". Given that all recent e500 are double precision capable, can we add an appropriate USE flag to gcc ebuild? Reproducible: Always
you can use EXTRA_ECONF in per-package env settings
I can use a lot of things, manual compilation included. :-) Considering that e500 constitute a visible part of deployed PPC machines and that it requires a complete rebuilding of the world (it has an incompatible FP architecture), USE flag is the least we can do.
@ppc: does this still make sense?
I think the problem with 'powerpc-linux-gnuspe' triplet is nobody (gcc, glibc, binutils) understands why it should have 'double' support. Something like 'crossdev powerpc-e500v2-linux-gnuspe' should force eclasses to pass that flags: https://gitweb.gentoo.org/repo/gentoo.git/tree/eclass/toolchain.eclass#n1134
A sudden activity on this bug report made me feel nostalgic. :-) Some e500v2 chips are probably still around, but Freescale had decided not to pursue the development of that ISA (switching back to normal PPC with all subsequent cores), so it's rather unlikely that people may need this in the future (which happens to be now).
(In reply to Alexander Dubov from comment #5) > A sudden activity on this bug report made me feel nostalgic. :-) > > Some e500v2 chips are probably still around, but Freescale had decided not > to pursue the development of that ISA (switching back to normal PPC with all > subsequent cores), so it's rather unlikely that people may need this in the > future (which happens to be now). Ack. Let's forget about it then. Thank you! :)
