Hi,
this PR is about rounding issue of sreal->integer conversion. I converted the
whole
code to sreals but the change seems bit too big and with no practical
improvement
to be justifiable, so this patch goes by simply fixing the template for current
roundoff
error. Hopefully counter code is now stable enough so we won't more random
flips in
GCC 8 release window.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
PR ipa/83179
* gcc.dg/ipa/inline-2.c: Fix template.
* gcc.dg/ipa/inline-3.c: Fix template.
Index: testsuite/gcc.dg/ipa/inline-2.c
===================================================================
--- testsuite/gcc.dg/ipa/inline-2.c (revision 257182)
+++ testsuite/gcc.dg/ipa/inline-2.c (working copy)
@@ -28,6 +28,6 @@ void foo (int invariant)
}
}
/* After inlining bar into foo, op2 is invariant within inner loop. */
-/* { dg-final { scan-ipa-dump "op2 change 10.000000. of time" "inline" } } */
+/* { dg-final { scan-ipa-dump "op2 change 9.990000. of time" "inline" } } */
/* After inlining bar into foo, op3 is invariant within both loops. */
/* { dg-final { scan-ipa-dump "op3 change 1.000000. of time" "inline" } } */
Index: testsuite/gcc.dg/ipa/inline-3.c
===================================================================
--- testsuite/gcc.dg/ipa/inline-3.c (revision 257182)
+++ testsuite/gcc.dg/ipa/inline-3.c (working copy)
@@ -21,4 +21,4 @@ int foo (int invariant)
}
-/* { dg-final { scan-ipa-dump "Scaling time by probability:0.100000" "inline"
} } */
+/* { dg-final { scan-ipa-dump "Scaling time by probability:0.099900" "inline"
} } */
