GMP 6.0.0 released

Torbjorn Granlund tg at
Wed Mar 26 18:14:10 UTC 2014

nisse at (Niels Möller) writes:

  I noticed that. Could be a bit confusing if different values are used in
  C and asm files, but no real problem, I hope.
I am not aware of any situation where different values are used and/or
could cause problems.

  And defining a new, more local, default in the assembly file using a
  particular threshold makes some sense.
  For most architectures, there's a top-level gmp-mparam.h,
  mpn/<arch>/gmp-mparam.h, but not for powerpc64. Maybe it should have
  one? Or are those files obsolete?

They will be used then our config.guess fails to identify the exact CPU.

For powerpc64, there is a mode64/gmp-mparam.h which will be used for any
build using that ABI.  And it's not kept up-to-date...

  E.g, x86_64/gmp-mparam.h carries a header
    /* AMD K8-K10 gmp-mparam.h -- Compiler/machine parameter header file.
    Copyright 1991, 1993, 1994, 2000-2010, 2012 Free Software Foundation, Inc.
  but then there are more recent x86_64/k8/gmp-mparam.h and
I am aware of this for x86_64.  It is not clear what a top-level files
should contain.  They are not normally used, except apparently under KVM
which by default suppresses the host CPU type.

We could generate a top-level file with a mean of the current CPUs'
values.  Or some weighted mean where the weight is determined by CPU
popularity and age.  :-)

Please encrypt, key id 0xC8601622

More information about the gmp-devel mailing list