Test failures on GNU/Linux with Athlon XP (32 bit) and GCC 4.3.2

Alexei Sheplyakov alexei.sheplyakov at gmail.com
Thu Sep 16 16:05:33 CEST 2010


Torbjorn Granlund <tg at gmplib.org> wrote:

> This looks awfully strange.  The toom22 code uses just assembly, and if
> toom22 fails, so should many other tests as well.

> If you run make check twice, so you get the same failures?

The set of failed test is the same every time, but the actual output is NOT
the same. Also sometimes t-div loops endlessly.

>  (I suspect  hardware issues.)

Given that the same problem exists with GCC 4.1.3 and 4.4.4, this
sounds convincing.
I'll try to reproduce the problem on a different machine, and will
report later on.

Best regards,

More information about the gmp-bugs mailing list