http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54886
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
AssignedTo|unassigned at gcc dot |rguenth at gcc dot gnu.org
|gnu.org |
--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-07
12:01:45 UTC ---
fooisl_ctx.c:158: isl_ctx freed, but some objects still reference it
this probably means some other data structure leaks:
==6257== 23,443 (48 direct, 23,395 indirect) bytes in 1 blocks are definitely
lo
st in loss record 778 of 781
==6257== at 0x4C2B14B: malloc (in
/usr/lib64/valgrind/vgpreload_memcheck-amd6
4-linux.so)
==6257== by 0x509087D: isl_pw_aff_alloc_size (in
/usr/lib64/libisl.so.10.0.0)
==6257== by 0x5090D3E: ??? (in /usr/lib64/libisl.so.10.0.0)
==6257== by 0x10651E9: wrap(isl_pw_aff*, unsigned int)
(graphite-sese-to-poly
.c:721)
==6257== by 0x106572B: extract_affine(scop*, tree_node*, isl_space*)
(graphite-sese-to-poly.c:833)
==6257== by 0x1065E22: build_loop_iteration_domains(scop*, loop*, int,
isl_set*, isl_set**) (graphite-sese-to-poly.c:1018)
==6257== by 0x1066D69: build_scop_iteration_domain(scop*)
(graphite-sese-to-poly.c:1411)
==6257== by 0x106B641: build_poly_scop(scop*) (graphite-sese-to-poly.c:3193)
==6257== by 0x10526F3: graphite_transform_loops() (graphite.c:287)
==6257== by 0xBEFE74: graphite_transforms() (tree-ssa-loop.c:258)
==6257== by 0x9C3EBE: execute_one_pass(opt_pass*) (passes.c:2335)
==6257== by 0x9C40B2: execute_pass_list(opt_pass*) (passes.c:2383)
and some more. I have a patch for the testcases at least.
