Re: PR11867, bignum operations

[Alan Modra <amodra at gmail dot com>]

On Tue, Aug 03, 2010 at 03:16:00PM -0400, DJ Delorie wrote:
> What was actually broken that triggered all this?

$ gas/as-new /src/cygnus/gas/testsuite/gas/all/octa.s
$ binutils/objdump -s a.out 

a.out:     file format elf32-m32c

Contents of section .data:
 0000 ffffddee bbcc99aa 77885566 3344ffff  ........w.Uf3D..
 0010 ffffddee bbcc99aa 77885566 34440000  ........w.Uf4D..

That's really mixed up, neither big nor little endian.
Expected little-endian results would be

 0000 ffff3344 55667788 99aabbcc ddeeffff  ..3DUfw.........
 0010 00003444 55667788 99aabbcc ddeeffff  ..4DUfw.........

But I'm really confused.  I'm looking at
M32C/80 Series
Software Manual
RENESAS 16/32-BIT SINGLE-CHIP
MICROCOMPUTER
M16C FAMILY / M32C/80 SERIES

NOP is 0xde according to that, but we generate 0x04.  I tried a few
other insns too.  None of them matched the manual.

-- 
Alan Modra
Australia Development Lab, IBM