On 2 January 2015 at 19:21, Ard Biesheuvel <ard.biesheu...@linaro.org> wrote: > I think the union type may have been a mistake to > begin with, because it introduces endianness dependencies that don't > actually exist in the code. It probably would have been cleaner if I > had defined the relation between VFP D-regs, words and bytes in terms > of shifts and masks instead.
Mmm, I think in retrospect this is right. Still, this patch does cause the kernel's self-tests on boot on a BE ppc64 host to pass, and looking through the code we haven't missed any accesses to .bytes or .words, so it looks good to me. Reviewed-by: Peter Maydell <peter.mayd...@linaro.org> -- PMM