From patchwork Sun Feb 27 12:13:52 2011
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: Avoid uninitialized padding bits in REAL_CST/CONST_DOUBLE (PR
middle-end/47903)
From: Jakub Jelinek
X-Patchwork-Id: 84686
Message-Id: <20110227121351.GC30899@tyan-ft48-01.lab.bos.redhat.com>
To: gcc-patches@gcc.gnu.org
Date: Sun, 27 Feb 2011 13:13:52 +0100
Hi!
In some places CONST_DOUBLE bits are compared as the HWIs, thus it is
undesirable to have uninitialized bits among them. At least on LP64 hosts
there are 32 bits of padding between uexp and sign fields.
Most of the places in gcc that work with REAL_VALUE_TYPE seem to actually
either clear the whole REAL_VALUE_TYPE by memset, or start by copying
to result one of the arguments. The exception seems to be real_arithmetic
for PLUS/MINUS/MULT/RDIV, where the result might be the same as either
of the operands, or might be completely different. Either real_arithmetic
in those 4 cases could compare result ptr with both operands and if it is
not equal to either of them, clear it, or the following patch adjusts
the callers where they do it.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2011-02-27 Jakub Jelinek
PR middle-end/47903
* fold-const.c (const_binop): Clear value first before
doing real_arithmetic into it.
* simplify-rtx.c (simplify_const_binary_operation): Likewise.
* builtins.c (fold_builtin_cbrt): Clear dconstroot first
before doing real_arithmetic into it.
(fold_builtin_modf): Clear frac first before doing real_arithmetic
into it.
Jakub
--- gcc/fold-const.c.jj 2011-02-15 15:25:53.000000000 +0100
+++ gcc/fold-const.c 2011-02-26 16:03:22.000000000 +0100
@@ -1163,6 +1163,8 @@ const_binop (enum tree_code code, tree a
else if (REAL_VALUE_ISNAN (d2))
return arg2;
+ /* Make sure to clear any padding. */
+ memset (&value, '\0', sizeof (value));
inexact = real_arithmetic (&value, code, &d1, &d2);
real_convert (&result, mode, &value);
--- gcc/simplify-rtx.c.jj 2010-12-02 11:51:31.000000000 +0100
+++ gcc/simplify-rtx.c 2011-02-26 16:06:38.000000000 +0100
@@ -3310,6 +3310,8 @@ simplify_const_binary_operation (enum rt
/* Inf * 0 = NaN plus exception. */
return 0;
+ /* Make sure to clear any padding. */
+ memset (&value, '\0', sizeof (value));
inexact = real_arithmetic (&value, rtx_to_tree_code (code),
&f0, &f1);
real_convert (&result, mode, &value);
--- gcc/builtins.c.jj 2011-02-02 16:30:55.000000000 +0100
+++ gcc/builtins.c 2011-02-26 16:02:49.000000000 +0100
@@ -7335,6 +7335,8 @@ fold_builtin_cbrt (location_t loc, tree
tree tree_root;
REAL_VALUE_TYPE dconstroot;
+ /* Make sure to clear any padding. */
+ memset (&dconstroot, '\0', sizeof (dconstroot));
real_arithmetic (&dconstroot, MULT_EXPR,
dconst_third_ptr (), dconst_third_ptr ());
dconstroot = real_value_truncate (TYPE_MODE (type), dconstroot);
@@ -9576,6 +9578,8 @@ fold_builtin_modf (location_t loc, tree
case rvc_normal:
/* Return (*arg1 = trunc(arg0), arg0-trunc(arg0)). */
real_trunc (&trunc, VOIDmode, value);
+ /* Make sure to clear any padding. */
+ memset (&frac, '\0', sizeof (frac));
real_arithmetic (&frac, MINUS_EXPR, value, &trunc);
/* If the original number was negative and already
integral, then the fractional part is -0.0. */