Re: [PATCH] Force rtl templates to be inlined

[Andi Kleen]

> I suspect the bulk of them currently are coming from the safe_as_a
> <rtx_insn *> calls within NEXT_INSN and PREV_INSN; do you happen to have
> information handy on that?

Yes that's right:

-   1.03%  lto1                    [.] bool 
is_a_helper<rtx_insn*>::test<rtx_def>(rtx_def*)                                 
                    ▒
   - bool is_a_helper<rtx_insn*>::test<rtx_def>(rtx_def*)                       
                                                                ▒
      - 92.20% bool is_a<rtx_insn*, rtx_def>(rtx_def*)                          
                                                                ▒
         - 98.53% rtx_insn* safe_as_a<rtx_insn*, rtx_def>(rtx_def*)             
                                                                ▒
            - 73.28% NEXT_INSN(rtx_insn const*)                                 
                                                                ▒