This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] Fix PR44828
- From: Richard Guenther <rguenther at suse dot de>
- To: gcc-patches at gcc dot gnu dot org
- Date: Tue, 6 Jul 2010 15:35:24 +0200 (CEST)
- Subject: [PATCH] Fix PR44828
Yet another broken frontend optimization. When narrowing a MULT_EXPR
we have to care for overflow as well. The fancy double-precision
check is needed to make the vectorizer pattern matching happy.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2010-07-06 Richard Guenther <rguenther@suse.de>
PR middle-end/44828
* convert.c (convert_to_integer): Watch out for overflowing
MULT_EXPR as well.
* gcc.c-torture/execute/pr44828.c: New testcase.
Index: gcc/testsuite/gcc.c-torture/execute/pr44828.c
===================================================================
*** gcc/testsuite/gcc.c-torture/execute/pr44828.c (revision 0)
--- gcc/testsuite/gcc.c-torture/execute/pr44828.c (revision 0)
***************
*** 0 ****
--- 1,19 ----
+ extern void abort (void);
+
+ static char
+ foo (char si1, char si2)
+ {
+ return si1 * si2;
+ }
+
+ int a = 0x105F61CA;
+
+ int
+ main (void)
+ {
+ int b = 0x0332F5C8;
+ if (foo (b, a) > 0)
+ abort ();
+ return 0;
+ }
+
Index: gcc/convert.c
===================================================================
*** gcc/convert.c (revision 161865)
--- gcc/convert.c (working copy)
*************** convert_to_integer (tree type, tree expr
*** 768,780 ****
|| ex_form == LSHIFT_EXPR
/* If we have !flag_wrapv, and either ARG0 or
ARG1 is of a signed type, we have to do
! PLUS_EXPR or MINUS_EXPR in an unsigned
! type. Otherwise, we would introduce
signed-overflow undefinedness. */
|| ((!TYPE_OVERFLOW_WRAPS (TREE_TYPE (arg0))
|| !TYPE_OVERFLOW_WRAPS (TREE_TYPE (arg1)))
&& (ex_form == PLUS_EXPR
! || ex_form == MINUS_EXPR)))
typex = unsigned_type_for (typex);
else
typex = signed_type_for (typex);
--- 768,786 ----
|| ex_form == LSHIFT_EXPR
/* If we have !flag_wrapv, and either ARG0 or
ARG1 is of a signed type, we have to do
! PLUS_EXPR, MINUS_EXPR or MULT_EXPR in an unsigned
! type in case the operation in outprec precision
! could overflow. Otherwise, we would introduce
signed-overflow undefinedness. */
|| ((!TYPE_OVERFLOW_WRAPS (TREE_TYPE (arg0))
|| !TYPE_OVERFLOW_WRAPS (TREE_TYPE (arg1)))
+ && ((TYPE_PRECISION (TREE_TYPE (arg0)) * 2u
+ > outprec)
+ || (TYPE_PRECISION (TREE_TYPE (arg1)) * 2u
+ > outprec))
&& (ex_form == PLUS_EXPR
! || ex_form == MINUS_EXPR
! || ex_form == MULT_EXPR)))
typex = unsigned_type_for (typex);
else
typex = signed_type_for (typex);