This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[RFA/testsuite] Disable loop unrolling for gcc.dg/pr34263.c

I set flag_unroll_loops = 1 in my backend for a test.  This test
failed, because it is checking for a loop cleanup optimization - but
tree level complete loop unrolling will remove the loop before we get
here, so there is nothing to clean up.

Is this reasonable to commit?  I didn't try to handle --target_board
unix/-funroll-loops which will put the multilib option after the
testcase option; that's something we've been having increasing
difficulty with, but I'm frankly not brave enough to change it.
I'm sure there'd be all kinds of fallout.

Daniel Jacobowitz

2009-04-23  Daniel Jacobowitz  <>

	* testsuite/gcc.dg/pr34263.c: Add -fno-unroll-loops.

Index: gcc/testsuite/gcc.dg/pr34263.c
--- gcc/testsuite/gcc.dg/pr34263.c	(revision 243180)
+++ gcc/testsuite/gcc.dg/pr34263.c	(working copy)
@@ -1,5 +1,5 @@
 /* { dg-do run } */
-/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-options "-O2 -fdump-tree-optimized -fno-unroll-loops" } */
 /* Same test as 990128-1.c.  */
 extern int printf (const char *,...);

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]