Using http://bitbucket.org/eigen/eigen/get/3.1.0-alpha1.tar.bz2 on Intel 11.1 I see a runtime failure in 513 - evaluators (Failed) The same test works fine for G++ 4.5.1. Details on the failing configuration follow. - Rhys -- ************************************************************ -- *** Eigen's unit tests configuration summary *** -- ************************************************************ -- -- Build type: Release -- Build site: setun -- Build string: linux-2.6.32-31-generic-g++icpc (icc) 11.1 20100806 copyright (c) 1985-2010 intel corporation. all rights reserved.-32bit -- Enabled backends: Cholmod, UmfPack, Qt4 support, fftw, OpenGL, GSL, -- Disabled backends: SuperLU, GoogleHash, Adolc, MPFR C++, -- Default order: Column-major -- Maximal matrix/vector size: 320 -- SSE2: Using architecture defaults -- SSE3: Using architecture defaults -- SSSE3: Using architecture defaults -- SSE4.1: Using architecture defaults -- SSE4.2: Using architecture defaults -- Altivec: Using architecture defaults -- ARM NEON: Using architecture defaults
Notice this is at -O3. Not sure what optimization level is supported for Intel. Feel free to tell me to re-run it at -O2 if you like.
While this of course needs to be fixed eventually, evaluators will not make it into 3.1. Would it make sense to disable this test for 3.1, or even remove this file and AssignEvaluator.h and CoreEvaluators.h temporarily and resurrect them when we branch 3.2? It could be quite confusing if we include code in Eigen 3.1 that isn't used anywhere.
Sounds like a plan.
I agree with that too
I resurrected the changes: https://bitbucket.org/eigen/eigen/changeset/0621fc0573b5/ changeset: 0621fc0573b5 user: ggael date: 2012-06-22 09:39:35 summary: resurrect expression evaluators
This unit test has been passing successfully for a while now.
-- 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/388.