Hi, As discussed at https://gcc.gnu.org/ml/gcc-patches/2014-10/msg00997.html, this is the one line patch which improves shrink wrapping for lo_sum.
OK? Bootstrapped and tested on x86_64 and built and tested for aarch64-elf with no regressions. It fixes shrink_wrap_symbol_ref_1.c which was failing with -mabi=ilp32. Thanks, Andrew ChangeLog: * shrink-wrap.c (move_insn_for_shrink_wrap): Allow LO_SUM also.
diff --git a/gcc/shrink-wrap.c b/gcc/shrink-wrap.c index aae6643..3392cdb 100644 --- a/gcc/shrink-wrap.c +++ b/gcc/shrink-wrap.c @@ -248,6 +248,7 @@ move_insn_for_shrink_wrap (basic_block bb, rtx_insn *insn, case SUBREG: case STRICT_LOW_PART: case PC: + case LO_SUM: /* Ok. Continue. */ break;