[PATCH] Record MEM stats from fold_buildN

[Andrew Pinski <pinskia at physics dot uc dot edu>]

If we build a tree because fold_binary returns 0, we just record the 
stat for that
memory location as fold_buildN but it would be better to record the 
call of
fold_buildN location instead.  This is like before were we did a "fold 
(buildN (",
we get the location of the buildN function call, now all we get is 
fold_buildN which
is not as useful.

OK? Bootstrapped and tested on powerpc-darwin with no regressions.

Thanks,
Andrew Pinski

ChangeLog:
	* tree.h (fold_build1): Change to macro and call fold_build1_stat.
	(fold_build2): Likewise.
	(fold_build3): Likewise.
	(fold_build1_stat): New function prototype.
	(fold_build2_stat): Likewise.
	(fold_build3_stat): Likewise.
	* fold-const.c (fold_build1): Rename to ..
	(fold_build1_stat): this.  Add MEM_STAT_DECL.  Pass the mem stats
	through to build1_stat.
	(fold_build2): Rename to ..
	(fold_build2_stat): this.  Add MEM_STAT_DECL.  Pass the mem stats
	through to build2_stat.
	(fold_build3): Rename to ..
	(fold_build3_stat): this.  Add MEM_STAT_DECL.  Pass the mem stats
	through to build3_stat.

Attachment: t.diff.txt
Description: Text document