Any Progress on s390x compile problem of gmp 4.2.4

[Torbjorn Granlund]

Rainer Dorsch <rdorsch at web.de> writes:

  Nevertheless, the problem seems still to be there and I run last week in it. I 
  googled and found the mailthread. So let me follow-up. This is the output of 
  config.guess:
  
  mclinx:firstsb /afs/bb/data/d3197/galaxy2/src/gmp-4.2.4+dfsg # ./config.guess
  s390x-ibm-linux
  mclinx:firstsb /afs/bb/data/d3197/galaxy2/src/gmp-4.2.4+dfsg #
  
We should try to make this work.  If somebody can respond to the
following questions, that would help.

Do these systems support a 32-bit ABI?  If they do, the compiler
should have an option controlling ABI.  Please try it, using something
trivial like hello.c.

What 64-bit multiply support do these system provide?  The 32-bit s390
systems have a signed 32x32->64 bit multiply.  GMP really needs a
64x64->128 bit multiply, preferably unsigned, to support the s390x
well.

Which compiler predefines are defined in 64-bit mode?  If there is an
32-bit ABI, what symbols are predefined when that ABI is chosen?

In longlong.h, we test the following predefines:

  __i370__
  __s390__
  __mvs__

I am not sure these symbols are the right ones, but they seem to work
for the 32-bit s390 systems.

-- 
Torbjörn