This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: inserting insn's into insn chain
- From: Per Fransson <per dot fransson dot 768 at student dot lu dot se>
- To: Richard Henderson <rth at redhat dot com>
- Cc: Per Fransson <per dot fransson dot 768 at student dot lu dot se>, gcc <gcc at gcc dot gnu dot org>
- Date: Tue, 21 May 2002 11:02:03 +0200
- Subject: Re: inserting insn's into insn chain
- References: <00ba01c1ff5d$6b033be0$9cf12fc2@sparta.afb.lu.se><20020519152626.D6055@redhat.com>
But emit_move_insn and emit_library_call seem to be inserting the insns
at the end of the chain. I'm working on an already fully generated
chain. Can't that be manipulated? Isn't that what some of the optimizing
passes do? My code gets called from toplev.c just before the
RTL-debugging info gets dumped (the one you get with the -dr option).
Boy, it's difficult to get the big picture here.
/Per
Richard Henderson wrote:
>On Sun, May 19, 2002 at 07:48:43PM +0200, Per Fransson wrote:
>
>>The store/restore code gets inserted into the RTL-code without problems.
>>Unfortunately it somehow gets removed before the assembler code is
>>generated.
>>
>
>It must be dead code then.
>
>>My code looks like this ...
>>
>
>What in the world are you doing? You should not be generating
>rtl at this level. You should be calling emit_move_insn, and
>emit_library_call.
>
>
>r~
>