This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Current state of MIPS16 support?
Hello Christopher,
On Wed, Apr 17, 2002 at 03:32:55PM -0700, Eric Christopher wrote:
> mips16 support was broken for quite some time. Recently a lot of bug
> fixes have gone into making it quite usable, though fairly unstable since
> it doesn't receive a lot of testing. Currently I test mips16 using a
> mips16 multilib in t-elf that isn't part of the regular set of multilibs
> built for mips-elf. Currently all of the bug fixes have only gone into
> the current trunk and not all of the work has gone into the branch
> (regression only fixes into the branch and mips16 has been broken since
> mid-99 so it wasn't as high a priority).
Thanks for the information.
I will do some testing in the next few weeks/months, as it
fits my schedule.
Can you give me more specific directions on how to build gcc with
the mips16 multilib? Configure for target mips-linux and
change t-elf like this?
MULTILIB_OPTIONS = mips16/no-mips16
MULTILIB_DIRNAMES = mips16 mips
MULTILIB_MATCHES =
> As far as linux support goes, support similar to that in t-elf would be
> necessary for mips16 stubs to exist in linux. AFAIK no one has tried
> using mips16 code in the kernel.
I currently won't try to use mips16 in the kernel, just for user space
applications. I asked on the linux-mips list, and the replies indicated
that it should work. There are a few places where the Linux kernel
interprets instructions (unaligned.c and traps.c/branch.c) which
might have to be changed, depending on what the user space code does.
Regards,
Johannes