> Hi,
> this patch fixes ICE while building Firefox (and probably xalancbmk, too)
> with -O3 -flto. I originally tested the whole patchset on several bigger apps
> including the inline heuristics change which teach it that thunks are very
> cheap.
> Mainline doesn't contain it that makes us to inline into thunks more heavilly
> and results in some suprises. I hope it is last one.
>
> Bootstrapped/regtested x86_64-linux, I am re-testing with last minute change
> and will commit afterwards.
>
> Honza
>
Hi,
this is variant of patch I comitted. It gets re-arranging of clone tree right
in function body saving and also fixes another ICE (probably unrelated) in
tree-inline.c when saving suddenly folds into direct call that is not in the
symbol table that can happen when we know we will redirect the call later.
Bootstrapped/regtested x86_64-linux.
Honza
* ipa-inline-transform.c (preserve_function_body_p): Look for
first non-thunk clone.
(save_function_body): Save into first non-thunk.
* lto-cgraph.c (lto_output_edge): When streaming thunk do not look
up call stmt id.
(lto_output_node): Inline thunks don't need body in every
partition.
* lto-streamer-in.c: Do not fixup thunk clones.
* cgraphclones.c (cgraph_node::create_edge_including_clone): Skip
thunks.
* tree-inline.c (copy_bb): Be prepared for target node to be new after
folding suceeds.
Index: cgraphclones.c
===================================================================
--- cgraphclones.c (revision 236317)
+++ cgraphclones.c (working copy)
@@ -771,33 +771,35 @@ cgraph_node::create_edge_including_clone
node = clones;
if (node)
while (node != this)
- {
- cgraph_edge *edge = node->get_edge (old_stmt);
+ /* Thunk clones do not get updated while copying inline function body.
*/
+ if (!node->thunk.thunk_p)
+ {
+ cgraph_edge *edge = node->get_edge (old_stmt);
- /* It is possible that clones already contain the edge while
- master didn't. Either we promoted indirect call into direct
- call in the clone or we are processing clones of unreachable
- master where edges has been removed. */
- if (edge)
- edge->set_call_stmt (stmt);
- else if (! node->get_edge (stmt))
- {
- edge = node->create_edge (callee, stmt, count, freq);
- edge->inline_failed = reason;
- }
+ /* It is possible that clones already contain the edge while
+ master didn't. Either we promoted indirect call into direct
+ call in the clone or we are processing clones of unreachable
+ master where edges has been removed. */
+ if (edge)
+ edge->set_call_stmt (stmt);
+ else if (! node->get_edge (stmt))
+ {
+ edge = node->create_edge (callee, stmt, count, freq);
+ edge->inline_failed = reason;
+ }
- if (node->clones)
- node = node->clones;
- else if (node->next_sibling_clone)
- node = node->next_sibling_clone;
- else
- {
- while (node != this && !node->next_sibling_clone)
- node = node->clone_of;
- if (node != this)
- node = node->next_sibling_clone;
- }
- }
+ if (node->clones)
+ node = node->clones;
+ else if (node->next_sibling_clone)
+ node = node->next_sibling_clone;
+ else
+ {
+ while (node != this && !node->next_sibling_clone)
+ node = node->clone_of;
+ if (node != this)
+ node = node->next_sibling_clone;
+ }
+ }
}
/* Remove the node from cgraph and all inline clones inlined into it.
Index: ipa-inline-transform.c
===================================================================
--- ipa-inline-transform.c (revision 236317)
+++ ipa-inline-transform.c (working copy)
@@ -506,6 +506,22 @@ save_inline_function_body (struct cgraph
/* first_clone will be turned into real function. */
first_clone = node->clones;
+
+ /* Arrange first clone to not be thunk as those do not have bodies. */
+ if (first_clone->thunk.thunk_p)
+ {
+ while (first_clone->thunk.thunk_p)
+ first_clone = first_clone->next_sibling_clone;
+ first_clone->prev_sibling_clone->next_sibling_clone
+ = first_clone->next_sibling_clone;
+ if (first_clone->next_sibling_clone)
+ first_clone->next_sibling_clone->prev_sibling_clone
+ = first_clone->prev_sibling_clone;
+ first_clone->next_sibling_clone = node->clones;
+ first_clone->prev_sibling_clone = NULL;
+ node->clones->prev_sibling_clone = first_clone;
+ node->clones = first_clone;
+ }
first_clone->decl = copy_node (node->decl);
first_clone->decl->decl_with_vis.symtab_node = first_clone;
gcc_assert (first_clone == cgraph_node::get (first_clone->decl));
@@ -514,7 +530,8 @@ save_inline_function_body (struct cgraph
first_clone. */
if (first_clone->next_sibling_clone)
{
- for (n = first_clone->next_sibling_clone; n->next_sibling_clone; n =
n->next_sibling_clone)
+ for (n = first_clone->next_sibling_clone; n->next_sibling_clone;
+ n = n->next_sibling_clone)
n->clone_of = first_clone;
n->clone_of = first_clone;
n->next_sibling_clone = first_clone->clones;
@@ -587,9 +604,10 @@ preserve_function_body_p (struct cgraph_
gcc_assert (symtab->global_info_ready);
gcc_assert (!node->alias && !node->thunk.thunk_p);
- /* Look if there is any clone around. */
- if (node->clones && !node->clones->thunk.thunk_p)
- return true;
+ /* Look if there is any non-thunk clone around. */
+ for (node = node->clones; node; node = node->next_sibling_clone)
+ if (!node->thunk.thunk_p)
+ return true;
return false;
}
Index: lto-cgraph.c
===================================================================
--- lto-cgraph.c (revision 236317)
+++ lto-cgraph.c (working copy)
@@ -259,7 +259,7 @@ lto_output_edge (struct lto_simple_outpu
streamer_write_gcov_count_stream (ob->main_stream, edge->count);
bp = bitpack_create (ob->main_stream);
- uid = (!gimple_has_body_p (edge->caller->decl)
+ uid = (!gimple_has_body_p (edge->caller->decl) || edge->caller->thunk.thunk_p
? edge->lto_stmt_uid : gimple_uid (edge->call_stmt) + 1);
bp_pack_enum (&bp, cgraph_inline_failed_t,
CIF_N_REASONS, edge->inline_failed);
@@ -398,7 +398,8 @@ lto_output_node (struct lto_simple_outpu
boundary_p = !lto_symtab_encoder_in_partition_p (encoder, node);
- if (node->analyzed && (!boundary_p || node->alias || node->thunk.thunk_p))
+ if (node->analyzed && (!boundary_p || node->alias
+ || (node->thunk.thunk_p && !node->global.inlined_to)))
tag = LTO_symtab_analyzed_node;
else
tag = LTO_symtab_unavail_node;
Index: lto-streamer-in.c
===================================================================
--- lto-streamer-in.c (revision 236317)
+++ lto-streamer-in.c (working copy)
@@ -953,7 +953,8 @@ fixup_call_stmt_edges (struct cgraph_nod
if (orig->clones)
for (node = orig->clones; node != orig;)
{
- fixup_call_stmt_edges_1 (node, stmts, fn);
+ if (!node->thunk.thunk_p)
+ fixup_call_stmt_edges_1 (node, stmts, fn);
if (node->clones)
node = node->clones;
else if (node->next_sibling_clone)
Index: tree-inline.c
===================================================================
--- tree-inline.c (revision 236317)
+++ tree-inline.c (working copy)
@@ -2063,7 +2063,7 @@ copy_bb (copy_body_data *id, basic_block
&& id->dst_node->definition
&& (fn = gimple_call_fndecl (stmt)) != NULL)
{
- struct cgraph_node *dest = cgraph_node::get (fn);
+ struct cgraph_node *dest = cgraph_node::get_create (fn);
/* We have missing edge in the callgraph. This can happen
when previous inlining turned an indirect call into a
