http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49948
Summary: ICE with -ftree-parallelize-loops: "address taken, but ADDRESSABLE bit not set" Product: gcc Version: 4.7.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: rtl-optimization AssignedTo: unassig...@gcc.gnu.org ReportedBy: arthur.j.odw...@gmail.com Created attachment 24894 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24894 Output of "ajo-gcc -O3 -ftree-parallelize-loops=2 -std=c99 -c test.c -v" This failure reproduces for me with svn revision 177081 (2011-08-02), and also with gcc 4.5.1. I'm on Ubuntu 10.10, x86-64 (-march=amdfam10). cat >test.c <<EOF extern int ej, *ep; int func_40() { int l_48, l_327 = 8; if (ej <= 7) { int *l_335[8]; for (int i = 0; i < 8; ++i) l_335[i] = &l_48; while (--l_327) { ej = 0; ep = l_335[0]; } return 0; } return (ep == &l_327); } EOF gcc -O3 -ftree-parallelize-loops=2 -std=c99 -c test.c test.c: In function ‘func_40’: test.c:13:5: error: address taken, but ADDRESSABLE bit not set l_48 cc1: note: in statement vect_cst_.15_1 = {&l_48, &l_48}; test.c:13:5: internal compiler error: verify_gimple failed With gcc 4.5.1, I get this ICE instead. Bug 28233 might be related: test.c: In function ‘func_40’: test.c:18:22: internal compiler error: in make_decl_rtl, at varasm.c:1347 This test case is reduced from the output of Csmith 2.1.0 (git hash 210eda7, https://github.com/Quuxplusone/csmith/), using the following command line: csmith --no-paranoid --no-longlong --pointers --arrays --jumps --no-consts --no-volatiles --no-checksum --divs --muls --bitfields --no-packed-struct -s 955105426