Re: [eigen] cwiseop_1 test failure

[ Thread Index | Date Index | More Archives ]

On Fri, 11 Jun 2010, Gael Guennebaud wrote:

ok, after some times I found that the error does not occur if I
compile with "-O2 -fno-guess-branch-probability" instead of "-O2"
only. Actually this optimization ( -f-guess-branch-probability) seems
to be related to the -fprofile-arcs option which is the object of the
aforementioned bug report. So I'll update it.

The guess-branch-probability switch certainly does something. In the current revision (c9ea79a108f2, after Benoit's big Dynamic --> -1 change), the number of test failures with gcc4.3 and sse drops down from 15 to 7 after the optimization is disabled. Specifically ...

Five tests fail regardless of whether it's enabled or not:
stable_norm_1, geo_alignedbox_7, stdvector_1, stdvector_overload_1, stdlist_1

Ten tests fail only when it's enabled (default Eigen release settings):
basicstuff_6, cwiseop_1, product_selfadjoint_1, product_selfadjoint_2, product_trsolve_6, geo_hyperplane_2, stdvector_5, stdvector_overload_5, stdlist_5, nesting_ops_4

Two tests fail only when it's disabled
("g++ -O2 -fno-guess-branch-probability"):
geo_parametrizedline_4, sparse_solvers_2

So just working around the GCC bug (if indeed it is one) by turning off this optimization will not work.


Mail converted by MHonArc 2.6.19+