[M3commit] CVS Update: cm3

[Jay Krell]

CVSROOT:	/usr/cvs
Changes by:	jkrell at birch.	10/11/20 13:48:03

Modified files:
	cm3/m3-sys/m3cc/gcc/gcc/m3cg/: parse.c 

Log message:
	restore TREE_SIDE_EFFECTS
	see recent change:
	Mon Nov 1 08:15:22 2010 MET (2 weeks, 5 days ago) by jkrell
	remove all but one use of TREE_SIDE_EFFECTS
	optimized m3core and libm3 are unchanged
	presumably unoptimized is also unchanged
	
	and maybe add a few more
	on all MODIFY_EXPR, BIND_EXPR, and calls
	and on volatile load/store (which we have very few of)
	
	It isn't automatic on those though?
	This is based on a quick perusal of the C frontend.
	I was really wondering what they do for nested functions,
	but ended up fixating on their add_stmt => add_to_statement_list_force
	vs. our add_stmt => add_to_statement_list which varies
	in that non-force doen't add if not TREE_SIDE_EFFECTS.
	
	I also don't know why we have TREE_USED in add_stmt. Left alone,
	just added a comment.
	
	Allow inlining in *modules* that have *no* nested functions,
	but not at all in *modules* that have *any* nested functions.
	I tried just nested functions and/or their parents, that's not enough.
	A little better but still not very good.
	
	It seems some of the previously broken cases for optimization
	aren't broken now. But still, overall, reenabling any/many
	of the few we have disabled, break somewhere.
	And some of this is probably related to type info or its absense.
	Continue punting for now.