Hi, This patch partially reverts r243782 where a return false was added expecting it to be a no-op. Detailed inspection shows this was not true. Despite no bug being identified following the change, removing the early return is likely to be safer than leaving it in place.
This is supported by the discussion here: https://gcc.gnu.org/ml/gcc/2017-02/msg00007.html Thanks, Matthew gcc/ * lra-constraints.c (simplify_operand_subreg): Remove early return false. --- gcc/lra-constraints.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/gcc/lra-constraints.c b/gcc/lra-constraints.c index 484a70d..0b7ae34 100644 --- a/gcc/lra-constraints.c +++ b/gcc/lra-constraints.c @@ -1607,7 +1607,8 @@ simplify_operand_subreg (int nop, machine_mode reg_mode) the memory. Typical case is when the index scale should correspond the memory. */ *curr_id->operand_loc[nop] = operand; - return false; + /* Do not return false here as the MEM_P (reg) will be processed + later in this function. */ } else if (REG_P (reg) && REGNO (reg) < FIRST_PSEUDO_REGISTER) { -- 2.2.1