PR 69400: Invalid 128-bit modulus result

As described in the PR, wi::divmod_internal was sign- rather than
zero-extending a modulus result in cases where the result has fewer
HWIs than the precision and the upper bit of the upper HWI was set.

This patch tries to make things more robust by getting wi_pack
to handle the canonicalisation step itself.

Tested on x86_64-linux-gnu.  I added tests to the wide-int
plugin since that seemed more direct.

gcc/
	PR tree-optimization/69400
	* wide-int.cc (wi_pack): Take the precision as argument and
	perform canonicalization here rather than in the callers.
	Use the main loop to handle all full-width HWIs.  Add a
	zero HWI if in_len isn't a full result.
	(wi::divmod_internal): Update accordingly.
	(wi::mul_internal): Likewise.  Simplify.

gcc/testsuite/
	PR tree-optimization/69400
	* gcc.dg/plugin/wide-int_plugin.c (test_wide_int_mod_trunc): New
	function.
	(plugin_init): Call it.
	* gcc.dg/torture/pr69400.c: New test.

From-SVN: r232817

gcc/ChangeLog

@@ -1,3 +1,13 @@
+2016-01-26  Richard Sandiford  <richard.sandiford@arm.com>
+
+	PR tree-optimization/69400
+	* wide-int.cc (wi_pack): Take the precision as argument and
+	perform canonicalization here rather than in the callers.
+	Use the main loop to handle all full-width HWIs.  Add a
+	zero HWI if in_len isn't a full result.
+	(wi::divmod_internal): Update accordingly.
+	(wi::mul_internal): Likewise.  Simplify.
+
 2016-01-25  Aditya Kumar  <aditya.k7@samsung.com>
 	    Sebastian Pop  <s.pop@samsung.com>
 

gcc/testsuite/ChangeLog

@@ -1,3 +1,11 @@
+2016-01-26  Richard Sandiford  <richard.sandiford@arm.com>
+
+	PR tree-optimization/69400
+	* gcc.dg/plugin/wide-int_plugin.c (test_wide_int_mod_trunc): New
+	function.
+	(plugin_init): Call it.
+	* gcc.dg/torture/pr69400.c: New test.
+
 2016-01-26  Christophe Lyon  <christophe.lyon@linaro.org>
 
 	* gcc.target/arm/pr68674.c: Check and use arm_fp effective target.

gcc/testsuite/gcc.dg/plugin/wide-int_plugin.c

@@ -36,11 +36,44 @@ test_wide_int_round_sdiv (void)
     abort ();
 }
 
+static void
+test_wide_int_mod_trunc (void)
+{
+  for (unsigned int i = 1; i < MAX_BITSIZE_MODE_ANY_INT; ++i)
+    {
+      if (wi::smod_trunc (wi::lshift (1, i + 1) - 3,
+			  wi::lshift (1, i) - 1)
+	  != wi::lshift (1, i) - 2)
+	abort ();
+      for (unsigned int base = 32; base <= MAX_BITSIZE_MODE_ANY_INT; base *= 2)
+	for (int bias = -1; bias <= 1; ++bias)
+	  {
+	    unsigned int precision = base + bias;
+	    if (i + 1 < precision && precision <= MAX_BITSIZE_MODE_ANY_INT)
+	      {
+		wide_int one = wi::uhwi (1, precision);
+		wide_int a = wi::lshift (one, i + 1) - 3;
+		wide_int b = wi::lshift (one, i) - 1;
+		wide_int c = wi::lshift (one, i) - 2;
+		if (wi::umod_trunc (a, b) != c)
+		  abort ();
+		if (wi::smod_trunc (a, b) != c)
+		  abort ();
+		if (wi::smod_trunc (-a, b) != -c)
+		  abort ();
+		if (wi::smod_trunc (a, -b) != c)
+		  abort ();
+	      }
+	  }
+    }
+}
+
 int
 plugin_init (struct plugin_name_args *plugin_info,
 	     struct plugin_gcc_version *version)
 {
   test_double_int_round_udiv ();
   test_wide_int_round_sdiv ();
+  test_wide_int_mod_trunc ();
   return 0;
 }

gcc/testsuite/gcc.dg/torture/pr69400.c

@@ -0,0 +1,20 @@
+/* { dg-do run { target int128 } } */
+
+typedef unsigned __int128 u128;
+
+u128 __attribute__((noinline, noclone))
+foo(void)
+{
+	u128 u = -2;
+	u %= 0xffffffffffffffffllu;
+	return u;
+}
+
+int
+main()
+{
+	u128 x = foo();
+	if (x != 0xfffffffffffffffellu)
+		__builtin_abort();
+	return 0;
+}

gcc/wide-int.cc

@@ -1214,30 +1214,32 @@ wi_unpack (unsigned HOST_HALF_WIDE_INT *result, const HOST_WIDE_INT *input,
     result[j++] = mask;
 }
 
-/* The inverse of wi_unpack.  IN_LEN is the the number of input
+/* The inverse of wi_unpack.  IN_LEN is the number of input
-   blocks.  The number of output blocks will be half this amount.  */
+   blocks and PRECISION is the precision of the result.  Return the
-static void
+   number of blocks in the canonicalized result.  */
-wi_pack (unsigned HOST_WIDE_INT *result,
+static unsigned int
+wi_pack (HOST_WIDE_INT *result,
 	 const unsigned HOST_HALF_WIDE_INT *input,
-	 unsigned int in_len)
+	 unsigned int in_len, unsigned int precision)
 {
   unsigned int i = 0;
   unsigned int j = 0;
+  unsigned int blocks_needed = BLOCKS_NEEDED (precision);
 
-  while (i + 2 < in_len)
+  while (i + 1 < in_len)
     {
-      result[j++] = (unsigned HOST_WIDE_INT)input[i]
+      result[j++] = ((unsigned HOST_WIDE_INT) input[i]
-	| ((unsigned HOST_WIDE_INT)input[i + 1]
+		     | ((unsigned HOST_WIDE_INT) input[i + 1]
-	   << HOST_BITS_PER_HALF_WIDE_INT);
+			<< HOST_BITS_PER_HALF_WIDE_INT));
       i += 2;
     }
 
   /* Handle the case where in_len is odd.   For this we zero extend.  */
   if (in_len & 1)
-    result[j++] = (unsigned HOST_WIDE_INT)input[i];
+    result[j++] = (unsigned HOST_WIDE_INT) input[i];
-  else
+  else if (j < blocks_needed)
-    result[j++] = (unsigned HOST_WIDE_INT)input[i]
+    result[j++] = 0;
-      | ((unsigned HOST_WIDE_INT)input[i + 1] << HOST_BITS_PER_HALF_WIDE_INT);
+  return canonize (result, j, precision);
 }
 
 /* Multiply Op1 by Op2.  If HIGH is set, only the upper half of the
@@ -1460,19 +1462,8 @@ wi::mul_internal (HOST_WIDE_INT *val, const HOST_WIDE_INT *op1val,
 	  *overflow = true;
     }
 
-  if (high)
+  int r_offset = high ? half_blocks_needed : 0;
-    {
+  return wi_pack (val, &r[r_offset], half_blocks_needed, prec);
-      /* compute [prec] <- ([prec] * [prec]) >> [prec] */
-      wi_pack ((unsigned HOST_WIDE_INT *) val,
-	       &r[half_blocks_needed], half_blocks_needed);
-      return canonize (val, blocks_needed, prec);
-    }
-  else
-    {
-      /* compute [prec] <- ([prec] * [prec]) && ((1 << [prec]) - 1) */
-      wi_pack ((unsigned HOST_WIDE_INT *) val, r, half_blocks_needed);
-      return canonize (val, blocks_needed, prec);
-    }
 }
 
 /* Compute the population count of X.  */
@@ -1847,8 +1838,7 @@ wi::divmod_internal (HOST_WIDE_INT *quotient, unsigned int *remainder_len,
   unsigned int quotient_len = 0;
   if (quotient)
     {
-      wi_pack ((unsigned HOST_WIDE_INT *) quotient, b_quotient, m);
+      quotient_len = wi_pack (quotient, b_quotient, m, dividend_prec);
-      quotient_len = canonize (quotient, (m + 1) / 2, dividend_prec);
       /* The quotient is neg if exactly one of the divisor or dividend is
 	 neg.  */
       if (dividend_neg != divisor_neg)
@@ -1859,8 +1849,7 @@ wi::divmod_internal (HOST_WIDE_INT *quotient, unsigned int *remainder_len,
 
   if (remainder)
     {
-      wi_pack ((unsigned HOST_WIDE_INT *) remainder, b_remainder, n);
+      *remainder_len = wi_pack (remainder, b_remainder, n, dividend_prec);
-      *remainder_len = canonize (remainder, (n + 1) / 2, dividend_prec);
       /* The remainder is always the same sign as the dividend.  */
       if (dividend_neg)
 	*remainder_len = wi::sub_large (remainder, zeros, 1, remainder,