CompFUSE / DCA

DCA++
BSD 3-Clause "New" or "Revised" License
36 stars 28 forks source link

Extensive test: analysis_dca_pp_up_down_sym_test fails #92

Open PDoakORNL opened 5 years ago

PDoakORNL commented 5 years ago

With gcc/5.3.0 and cuda/8.0.61 the test fails.

This may be simply a case of an overly optimistic epsilon in the test.

    start reading function : leading-eigenvalues
    start reading function : leading-eigenvectors
    start reading function : leading-symmetry-decomposition../test/system-level/analysis/dca_pp_up_down_sym/analysis_dca_pp_up_down_sym_test.cpp:104: Failure
The difference between std::abs(leading_symmetry_decomposition_check(i).real()) and std::abs(leading_symmetry_decomposition(i).real()) is 1.2149403653259543e-10, which exceeds 1.e-10, where
std::abs(leading_symmetry_decomposition_check(i).real()) evaluates to 1.7234960465399499e-11,
std::abs(leading_symmetry_decomposition(i).real()) evaluates to 1.3872899699799492e-10, and
1.e-10 evaluates to 1e-10.
../test/system-level/analysis/dca_pp_up_down_sym/analysis_dca_pp_up_down_sym_test.cpp:104: Failure
The difference between std::abs(leading_symmetry_decomposition_check(i).real()) and std::abs(leading_symmetry_decomposition(i).real()) is 1.2257347761448061e-10, which exceeds 1.e-10, where
std::abs(leading_symmetry_decomposition_check(i).real()) evaluates to 1.7574631802727578e-11,
std::abs(leading_symmetry_decomposition(i).real()) evaluates to 1.401481094172082e-10, and
1.e-10 evaluates to 1e-10.
ubulling commented 5 years ago

I think you're right. We had to increase the tolerance before.