[PATCH] Invalid constraints on SPARC f?tox and fxto?

[Andrew Over <andrew dot over at cs dot anu dot edu dot au>]

Hi All.

I've recently been looking into gcc bug 10904, which involves gcc
generating instructions such as:

  fdtox %f8,%f7

on SPARC64 [1].

Sun's assembler chokes on this construct, while gas silently accepts it
(and according to Sun's disassembler, the %f7 is encoded as %f38).  One
consequence of this is that when using gas with gcc on sparc64, the
(broken) assembly is silently misassembled.

I've attached a patch against binutils 2.14 which I believe tightens up
these constraints.

Please CC me as I'm not on the list.

Cheers,
--Andrew

[1] The V9 arch manual I'm looking at doesn't seem to claim any alignment
    constraints on rd for f?tox, although Sun's assembler seems to
    believe otherwise.

Attachment: binutils.patch
Description: Text document