The following part of r126198 causes a 5% slowdown of tramp3d for the
non-leafify
tests (with leafify it makes it run faster):
2007-07-02 Richard Guenther <[EMAIL PROTECTED]>
* fold-const.c (fold_convert): Revert fix for PR15988.
* tree-inline.c (setup_one_parameter): Instead fix it here by
using fold_build1 instead of fold_convert and checking for
error_mark_node. Convert only if the conversion is necessary.
Index: fold-const.c
===================================================================
--- fold-const.c (revision 126197)
+++ fold-const.c (revision 126198)
@@ -2262,9 +2262,7 @@ fold_convert (tree type, tree arg)
|| TREE_CODE (orig) == ERROR_MARK)
return error_mark_node;
- if (TYPE_MAIN_VARIANT (type) == TYPE_MAIN_VARIANT (orig)
- || lang_hooks.types_compatible_p (TYPE_MAIN_VARIANT (type),
- TYPE_MAIN_VARIANT (orig)))
+ if (TYPE_MAIN_VARIANT (type) == TYPE_MAIN_VARIANT (orig))
return fold_build1 (NOP_EXPR, type, arg);
switch (TREE_CODE (type))
Index: tree-inline.c
===================================================================
--- tree-inline.c (revision 126197)
+++ tree-inline.c (revision 126198)
@@ -1278,10 +1278,15 @@ setup_one_parameter (copy_body_data *id,
tree init_stmt;
tree var;
tree var_sub;
- tree rhs = value ? fold_convert (TREE_TYPE (p), value) : NULL;
+ tree rhs = value;
tree def = (gimple_in_ssa_p (cfun)
? gimple_default_def (id->src_cfun, p) : NULL);
+ if (value
+ && value != error_mark_node
+ && !useless_type_conversion_p (TREE_TYPE (p), TREE_TYPE (value)))
+ rhs = fold_build1 (NOP_EXPR, TREE_TYPE (p), value);
+
/* If the parameter is never assigned to, has no SSA_NAMEs created,
we may not need to create a new variable here at all. Instead, we may
be able to just use the argument value. */
I am investigating why.
--
Summary: [4.3 Regression] r126198 causes tramp3d slowdown w/o
leafify
Product: gcc
Version: 4.3.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: middle-end
AssignedTo: rguenth at gcc dot gnu dot org
ReportedBy: rguenth at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32624
