gmp-5.1.3: architecture detection failure?

Niels Möller nisse at lysator.liu.se
Fri Nov 22 07:19:47 UTC 2013


Yves Caniou <yves.caniou at ens-lyon.fr> writes:

> - Is it normal that a Intel(R) Xeon(R) CPU W3670  3.20GHz is recognized as a 
> corei7? 

Don't know.

Most of the time, this fine-grained configuration is only for
performance tuning, and configuring with a slightly different flavor
should still produce a working library. Unless there are any relevant
instructions which are available on corei7 but not on the actual cpu?

> - And do you consider that `configfsf.guess` and `config.guess` not returning 
> the same output, as it said it would in the reportbug FAQ, is not an error?

Different output is normal. E.g., on my office machine, I get

  ~/hack/gmp$ ./configfsf.guess
  x86_64-unknown-linux-gnu
  ~/hack/gmp$ ./config.guess
  coreisbr-unknown-linux-gnu

The more worrying problem with the report was the mpc testsuite failure.
One possibility is that there's a bug in some corei7-specific assembly
file in gmp. Or it could be an bug in mpc. Or it could be a problem in
the cpu detection (but I'm not sure what's expected there). To address
this, you'd need to first isolate the problem to a smaller test program,
with only a small number of gmp calls, and no dependency on mpc. And in
that process you will likely find out if the problem is in mpc or in
gmp.

Regards,
/Niels

-- 
Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
Internet email is subject to wholesale government surveillance.


More information about the gmp-bugs mailing list