The test occasionally fails, but apparently not depending on the random-seed. Maybe there is a race condition somewhere. To reproduce, build the test and run this from the build-directory: while ./unsupported/test/cxx11_tensor_reduction s1; do echo " Success"; done after 1-10 tries this results in: > Difference too large wrt tolerance 0.0100021, relative error is: -nan > Test ( test_simple_reductions<Eigen::half,ColMajor>() ) failed in ../unsupported/test/cxx11_tensor_reduction.cpp (107) > verifyIsApprox(result(i, j), prod) > Stack: > - ( test_simple_reductions<Eigen::half,ColMajor>() ) > - cxx11_tensor_reduction > - Seed: 1 The 's1' has no effect here, this is just to illustrate the non-deterministic nature of the problem. Maybe there is a race condition in the code.
The reason for the non-deterministic behavior is Bug 1602.
-- GitLab Migration Automatic Message -- This bug has been migrated to gitlab.com's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.com/libeigen/eigen/issues/1597.