[PATCH] [lld] [mach-o] make lld respect alignment constraints more

[kledzik at apple.com]

How come in the ArchHandler_x86_64 you use:
    write32(loc, value, false);
but inside ArchHandler_arm64 you use:
    write32(loc, value, _isBig);

Seems like each arch has a fixed endianness.  Given that, 
    write32(loc, value, false);
could be optimized to:
    writeLittle32(loc, value); // or use little32_t directly?

Have you looked at the codegen of this verse the old 'swap' parameter.  Just to make sure the compiler really is optimizing what you think it should be?

http://reviews.llvm.org/D5811