https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71292
--- Comment #1 from kugan at gcc dot gnu.org --- (In reply to ktkachov from comment #0) > An aarch64-none-linux-gnu bootstrap with an in-tree mpfr fails with an ICE: > exp_2.i: In function ‘fn1’: > exp_2.i:4:6: internal compiler error: Segmentation fault > void fn1() { > ^~~ > 0xb1f8cf crash_signal > $SRC/gcc/toplev.c:333 > 0x89bcb9 bb_seq_addr > $SRC/gcc/gimple.h:1655 > 0x89bcb9 gsi_start_bb > $SRC/gcc/gimple-iterator.h:129 > 0x89bcb9 gsi_for_stmt(gimple*) > $SRC/gcc/gimple-iterator.c:617 > 0xcbbeba insert_stmt_after > $SRC/gcc/tree-ssa-reassoc.c:1323 > 0xcbd67a build_and_add_sum > $SRC/gcc/tree-ssa-reassoc.c:1392 > 0xcbf34f rewrite_expr_tree_parallel > $SRC/gcc/tree-ssa-reassoc.c:4128 > 0xcc8b95 reassociate_bb > $SRC/gcc/tree-ssa-reassoc.c:5339 > 0xcc8ad7 reassociate_bb > $SRC/gcc/tree-ssa-reassoc.c:5391 > 0xccb523 do_reassoc > $SRC/gcc/tree-ssa-reassoc.c:5505 > 0xccb523 execute_reassoc > $SRC/gcc/tree-ssa-reassoc.c:5592 > 0xccb523 execute > $SRC/gcc/tree-ssa-reassoc.c:5631 > Please submit a full bug report, > with preprocessed source if appropriate. > Please include the complete backtrace with any bug report. > See <http://gcc.gnu.org/bugs.html> for instructions. > > The reduced testcase for that reproducible with trunk at: > gcc version 7.0.0 20160526 > is: > > unsigned long a; > long b, d; > int c; > void fn1() { > unsigned long e = a + c; > b = d + e + a + 8; > } > > compile with -O2. > Compiling with -fno-tree-reassoc doesn't ICE It looks like dup of PR71252. does the patch at help? https://gcc.gnu.org/ml/gcc-patches/2016-05/msg02076.html