This is the mail archive of the gcc@gcc.gnu.org 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]

This was fun to track down



Mon Sep 22 22:18:48 1997  David S. Miller  <davem@tanya.rutgers.edu>

	* expmed.c (expand_divmod): If compute_mode is not the same as
	mode, handle the case where convert_modes() causes op1 to no
	longer be a CONST_INT.

--- expmed.c.~1~	Sun Sep  7 18:10:31 1997
+++ expmed.c	Mon Sep 22 22:18:27 1997
@@ -2831,6 +2831,14 @@
     {
       op0 = convert_modes (compute_mode, mode, op0, unsignedp);
       op1 = convert_modes (compute_mode, mode, op1, unsignedp);
+
+      /* convert_modes() may have tossed op1 into a register, so we
+	 absolutely must recompute the following.  */
+      op1_is_constant = GET_CODE (op1) == CONST_INT;
+      op1_is_pow2 = (op1_is_constant
+		     && ((EXACT_POWER_OF_2_OR_ZERO_P (INTVAL (op1))
+			  || (! unsignedp
+			      && EXACT_POWER_OF_2_OR_ZERO_P (-INTVAL (op1))))));
     }
 
   /* If one of the operands is a volatile MEM, copy it into a register.  */


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