reload_cse_noop_set_p(): invoking rtx_equal_for_cselib_p() is not enough; if

it reports the dest and source parts of the set insn are identical, we need
to check whether they are volatile memory references, and return 0 (not a
noop) in that case.

This allows `*volatile_ptr = *volatile_ptr;' constructs to no longer be
incorrectly optimized away on platforms which can perform memory to memory
transfers in a single instruction (i.e. m68k and vax).

Found the hard way on vax, where some systems need to frob the interrupt
stack early, before being able to handle faults, and a gcc3-built kernel
would have this operation optimized away, leading to a quick system reset.

1 files changed, 13 insertions, 1 deletions

diff --git a/gnu/usr.bin/gcc/gcc/reload1.c b/gnu/usr.bin/gcc/gcc/reload1.c
index 5f8102c8821..2cc4c3f7150 100644
--- a/gnu/usr.bin/gcc/gcc/reload1.c
+++ b/gnu/usr.bin/gcc/gcc/reload1.c
@@ -8059,7 +8059,19 @@ static int
 reload_cse_noop_set_p (set)
      rtx set;
 {
-  return rtx_equal_for_cselib_p (SET_DEST (set), SET_SRC (set));
+  rtx dest, src;
+
+  dest = SET_DEST (set);
+  src = SET_SRC (set);
+
+  if (! rtx_equal_for_cselib_p (dest, src))
+    return 0;
+
+  if ((GET_CODE (dest) == MEM && MEM_VOLATILE_P (dest))
+      || (GET_CODE (src) == MEM && MEM_VOLATILE_P (src)))
+    return 0;
+
+  return 1;
 }
 
 /* Try to simplify INSN.  */