Re: Why doesn't -fomit-frame-pointer work(very well)?

[Alan Lehotsky <apl at alum dot mit dot edu>]

At 6:27 PM -0500 4/4/02, Peter Barada wrote:

> >On Wed, Apr 03, 2002 at 06:01:40PM -0500, Peter Barada wrote:
>>> So it looks like it tried to eliminate the frame pointer and then
>>> failed.
>>
>>Hum.  I suspect that reload doesn't like %sp being modifed
>>and used within the same insn.  In fact, IIRC this is illegal
>>on ARM, so there may well be code in there to prevent this.
>
>I didn't find any explicit code to disallow modifying %sp in an
>instruction that refers to it.  After adding in ARG_POINTER_REGNUM to
>mimic what others have done (i386, h8300, etc), I have in 3.0.4(for
>the non-FPA case):


	I'm pretty sure that there is code in register elimination that won't eliminate if
	there's a change-and-use in the same instruction.

	This is because the semantics of that are machine-dependent (you might have a code pattern that
	knows it's safe to generate

		move -(sp),4(sp)

	because you know that the src address is evaluated before the stack is adjusted).

	I remember running into this before on a machine that only had 3 registers capable of indirection
	(and only two of them allowed any displacement offset!)  Making register elimination work was
	particularly critical!

	....	
	Okay, rather than being cryptic, I went and looked at the source code.....

	In reload1.c  elimination_effects() will stumble on things like push instructions that try to eliminate
	into SP-relative addresses.

	I fixed this in the case SET: leg of the switch stmt by recursively calling elimination_effects() on
	the SET_SRC with something like....

		mem_mode = GET_MODE (SET_DEST (x));
		elimination_effects (SET_SRC (x),
					(push_operand (SET_DEST (x), mem_mode) ? mem_mode : 0));

	instead of the

		elimination_effects (SET_SRC (x), 0);

	that's normally there.
-- 
------------------------------------------------------------------------

		    Quality Software Management
		http://home.earthlink.net/~qsmgmt
			apl@alum.mit.edu
			(978)287-0435 Voice
			(978)808-6836 Cell
			(978)287-0436 Fax

	Software Process Improvement and Management Consulting
	     Language Design and Compiler Implementation