* [PATCH] Fix PR80167
@ 2017-03-24 12:28 Richard Biener
0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2017-03-24 12:28 UTC (permalink / raw)
To: gcc-patches
Seems we have simply forgotten about the case of replacing with a
default-def.
Bootstrapped / tested on x86_64-unknown-linux-gnu, applied.
Richard.
2017-03-24 Richard Biener <rguenther@suse.de>
PR tree-optimization/80167
* graphite-isl-ast-to-gimple.c
(translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
properly.
(translate_isl_ast_to_gimple::get_rename): Likewise.
* gcc.dg/graphite/pr80167.c: New testcase.
Index: gcc/graphite-isl-ast-to-gimple.c
===================================================================
*** gcc/graphite-isl-ast-to-gimple.c (revision 246437)
--- gcc/graphite-isl-ast-to-gimple.c (working copy)
*************** bool translate_isl_ast_to_gimple::
*** 1123,1128 ****
--- 1123,1131 ----
is_valid_rename (tree rename, basic_block def_bb, basic_block use_bb,
phi_node_kind phi_kind, tree old_name, basic_block old_bb) const
{
+ if (SSA_NAME_IS_DEFAULT_DEF (rename))
+ return true;
+
/* The def of the rename must either dominate the uses or come from a
back-edge. Also the def must respect the loop closed ssa form. */
if (!is_loop_closed_ssa_use (use_bb, rename))
*************** get_rename (basic_block new_bb, tree old
*** 1178,1183 ****
--- 1181,1187 ----
basic_block bb = gimple_bb (SSA_NAME_DEF_STMT (rename));
if (is_valid_rename (rename, bb, new_bb, phi_kind, old_name, old_bb)
&& (phi_kind == close_phi
+ || ! bb
|| flow_bb_inside_loop_p (bb->loop_father, new_bb)))
return rename;
return NULL_TREE;
Index: gcc/testsuite/gcc.dg/graphite/pr80167.c
===================================================================
*** gcc/testsuite/gcc.dg/graphite/pr80167.c (nonexistent)
--- gcc/testsuite/gcc.dg/graphite/pr80167.c (working copy)
***************
*** 0 ****
--- 1,24 ----
+ /* { dg-do compile } */
+ /* { dg-options "-O2 -floop-nest-optimize" } */
+
+ typedef struct
+ {
+ short a;
+ short b;
+ short c;
+ } d;
+ extern d e[];
+ int f[8];
+ void
+ g (d *i)
+ {
+ int h = 0;
+ for (; h < 28; h++)
+ e[h].a = e[h].b = i[h].a;
+ h = 0;
+ for (; h < 8; h++)
+ f[h] = i[h].b + i[h].c;
+ h = 0;
+ for (; h < 8; h++)
+ f[h] = i[h].b;
+ }
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2017-03-24 12:14 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-03-24 12:28 [PATCH] Fix PR80167 Richard Biener
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).