Fix PR48622 (lto ICE, lto bootstrap)

Fix PR48622 (lto ICE, lto bootstrap)

[Michael Matz]

Hi,

since r172430 lto bootstrap is broken, as well as the attached testcase 
(pr48622) and cpu2006 compilation (pr48645).  The inline summary writer 
used a different order for size and time than the reader expected.

I've committed the below patch as obvious (r172603) after verifying that 
lto bootstrap passes the place of the ICE (and that the testcase is 
fixed).


Ciao,
Michael.

	PR tree-optimization/48622
	PR lto/48645
	* ipa-inline-analysis.c (inline_read_summary): Read size/time
	in same order as they're written.

testsuite/
	* gcc.dg/lto/pr48622_0.c: New test.

Index: ipa-inline-analysis.c
===================================================================
--- ipa-inline-analysis.c	(Revision 172602)
+++ ipa-inline-analysis.c	(Arbeitskopie)
@@ -612,10 +612,10 @@ inline_read_summary (void)
 
 	      info->estimated_stack_size
 	        = info->estimated_self_stack_size = lto_input_uleb128 (ib);
-	      info->time = info->self_time = lto_input_uleb128 (ib);
-	      info->time_inlining_benefit = lto_input_uleb128 (ib);
 	      info->size = info->self_size = lto_input_uleb128 (ib);
 	      info->size_inlining_benefit = lto_input_uleb128 (ib);
+	      info->time = info->self_time = lto_input_uleb128 (ib);
+	      info->time_inlining_benefit = lto_input_uleb128 (ib);
 	      info->estimated_growth = INT_MIN;
 
 	      bp = lto_input_bitpack (ib);
Index: testsuite/gcc.dg/lto/pr48622_0.c
===================================================================
--- testsuite/gcc.dg/lto/pr48622_0.c	(Revision 0)
+++ testsuite/gcc.dg/lto/pr48622_0.c	(Revision 0)
@@ -0,0 +1,93 @@
+/* { dg-lto-do link } */
+/* { dg-lto-options {{-O -flto -finline-small-functions -fno-early-inlining}} } */
+
+void abort(void);
+void exit(int);
+
+typedef unsigned int u8 __attribute__ ((mode (QI)));
+
+u8
+ashift_qi_0 (u8 n)
+{
+}
+
+u8
+ashift_qi_2 (u8 n)
+{
+}
+
+u8
+ashift_qi_3 (u8 n)
+{
+}
+
+u8
+ashift_qi_4 (u8 n)
+{
+}
+
+u8
+ashift_qi_5 (u8 n)
+{
+}
+
+u8
+ashift_qi_6 (u8 n)
+{
+}
+
+u8
+ashift_qi_7 (u8 n)
+{
+}
+
+u8
+lshiftrt_qi_0 (u8 n)
+{
+}
+
+u8
+lshiftrt_qi_4 (u8 n)
+{
+}
+
+u8
+lshiftrt_qi_5 (u8 n)
+{
+}
+
+u8
+lshiftrt_qi_7 (u8 n)
+{
+}
+
+int
+main ()
+{
+  if (ashift_qi_0 (0xff) != (u8) ((u8) 0xff << 0))
+    abort ();
+  if (ashift_qi_1 (0xff) != (u8) ((u8) 0xff << 1))
+    abort ();
+  if (ashift_qi_2 (0xff) != (u8) ((u8) 0xff << 2))
+    abort ();
+  if (ashift_qi_3 (0xff) != (u8) ((u8) 0xff << 3))
+    abort ();
+  if (ashift_qi_4 (0xff) != (u8) ((u8) 0xff << 4))
+    abort ();
+  if (ashift_qi_5 (0xff) != (u8) ((u8) 0xff << 5))
+    abort ();
+  if (ashift_qi_6 (0xff) != (u8) ((u8) 0xff << 6))
+    abort ();
+  if (ashift_qi_7 (0xff) != (u8) ((u8) 0xff << 7))
+    abort ();
+  if (lshiftrt_qi_0 (0xff) != (u8) ((u8) 0xff >> 0))
+    abort ();
+  if (lshiftrt_qi_4 (0xff) != (u8) ((u8) 0xff >> 4))
+    abort ();
+  if (lshiftrt_qi_5 (0xff) != (u8) ((u8) 0xff >> 5))
+    abort ();
+  if (lshiftrt_qi_7 (0xff) != (u8) ((u8) 0xff >> 7))
+    abort ();
+  exit (0);
+}
+

Re: Fix PR48622 (lto ICE, lto bootstrap)

[Jan Hubicka]

> Hi,
> 
> since r172430 lto bootstrap is broken, as well as the attached testcase 
> (pr48622) and cpu2006 compilation (pr48645).  The inline summary writer 
> used a different order for size and time than the reader expected.
> 
> I've committed the below patch as obvious (r172603) after verifying that 
> lto bootstrap passes the place of the ICE (and that the testcase is 
> fixed).

Oops, thanks!  I wonder why this did not wreck the LTO benchmark results though ;)

Honza