the current graphite branch segfaults & ICEs on several files from CP2K. Compiling with '-O2 -ffast-math -funroll-loops -ftree-vectorize -march=native -fgraphite -fgraphite-identity'
To reproduce get http://www.pci.uzh.ch/vandevondele/tmp/CP2K_2008_12_03.tgz untar and type make at least on the following files: mltfftsg.F mltfftsg_tools.F harris_force_types.F ai_moments.F cp_units.F ps_wavelet_util.F cp_fm_basic_linalg.F ps_wavelet_kernel.F ai_angmom.F distribution_optimize.F qs_util.F atom_fit.F atom_operators.F dynamical_coeff_types.F pw_spline_utils.F pw_poisson_types.F xc_derivative_set_types.F cp_ddapc_methods.F qs_all_potential.F qs_operators_ao.F I have ICEs. The first one is: ai_moments.F: In function cossin: ai_moments.F:44: internal compiler error: Segmentation fault Please submit a full bug report, with preprocessed source if appropriate. See <http://gcc.gnu.org/bugs.html> for instructions. make: *** [ai_moments.o] Error The same files compile fine with '-O0 -fgraphite -fgraphite-identity' so I guess it is related to applying graphite to optimized code (or the other way around). -- Summary: [graphite] several ICEs with CP2K Product: gcc Version: 4.4.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: middle-end AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: jv244 at cam dot ac dot uk http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38431