Fix math-test false positive
commitd02e524326c06cb33ac7fef3130d423f30771d4f
authorRoland Schulz <roland.schulz@intel.com>
Fri, 16 Sep 2016 00:45:33 +0000 (15 17:45 -0700)
committerErik Lindahl <erik.lindahl@gmail.com>
Sun, 18 Sep 2016 14:53:20 +0000 (18 16:53 +0200)
tree6bc08687a64070ec2461187f2b46d676426e6314
parentde644b90c81c12a4cb27b9ac2b57a4151711ab1b
Fix math-test false positive

Depending on the accuracy of the floating point division, the
input of the test function could be 1ulp too large or too small.
If it was too large the result of the test function wasn't
within 4ulp and the test failed.

Change-Id: Ia322b136d1db9b25c7c733e4067f927a4c77e372
src/gromacs/math/tests/functions.cpp