gmp-6.1.0 on Cygwin

Torbjörn Granlund tg at gmplib.org
Mon Nov 30 20:03:50 UTC 2015


ASSI <Stromeko at nexgo.de> writes:

  
  I've recently updated to gmp-6.1.0 on Cygwin and a few users have reported 
  problems with various bits and pieces of gcc segfaulting.  In all those 
  cases, going back to gmp-6.0.0a got things back to working.  Two problems 
  have been observed on Broadwell CPU (i5-5300U and i7-5600U) and all on 64bit 
  (x86_64).  Looking at the release notes the only thing that stands out is 

"All on 64bit"?  What does it mean?

Is the problem observed specific to Broadwell or are all x86_64 systems
affected?

Two problems?  Do you have specifics?

  the ADX instruction support for Broadwell/Skylake, which incidentally isn't 
  used in 32bit when a fat binary is built (at least my interpretation of 
  fat.c is that Broadwell/Skylake have no special treatment for 32bit, which 
  is a bit peculiar -- I'd think they should at least get listed along with 
  Haswell).  I can't test on either architecture myself.
  
  Since I don't want to downrev gmp for everybody, is there an easy way to 
  skip that bit of code when building gmp?  That would also provide an 
  ^^^^^^^^^^^^^
  indication of whether that is the right spot to further investigate where 
  the problems in the applications come from.
  
Skip what bit of code?

You don't tell us much about what your builds look like.

Are you using a fat build?  A 64-bit or a 32-bit build?

Please see our bug reporting manual page for information of what we
need.  This will speed thing up.

-- 
Torbjörn
Please encrypt, key id 0xC8601622


More information about the gmp-discuss mailing list