[4.7 PATCH] Optimize snprintf (buf, 5, "abcd") into memcpy (PR middle-end/47917)

[Jakub Jelinek <jakub at redhat dot com>]

Hi!

This patch optimizes snprintf with constant second argument
and format string without any % chars or "%s" if the second
argument is known to be greater than strlen of the string.

We could optimize also the case where it is smaller by emitting
either strcpy from a modified string (but then we risk .rodata growth),
or by doing memcpy on length - 1 followed by a single byte zero
store at the end of buffer, but then it might grow code size,
so for now I've left those cases out.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for 4.7?

2011-02-28  Jakub Jelinek  <jakub@redhat.com>

	PR middle-end/47917
	* builtins.c (fold_builtin_snprintf): New function.
	(fold_builtin_3): Call it for BUILT_IN_SNPRINTF.
	(fold_builtin_4): Likewise.

	* gcc.c-torture/execute/pr47917.c: New test.
	* gcc.dg/pr47917.c: New test.

--- gcc/builtins.c.jj	2011-02-26 16:02:49.000000000 +0100
+++ gcc/builtins.c	2011-02-28 14:00:04.000000000 +0100
@@ -191,6 +191,7 @@ static tree fold_builtin_strncat (locati
 static tree fold_builtin_strspn (location_t, tree, tree);
 static tree fold_builtin_strcspn (location_t, tree, tree);
 static tree fold_builtin_sprintf (location_t, tree, tree, tree, int);
+static tree fold_builtin_snprintf (location_t, tree, tree, tree, tree, int);
 
 static rtx expand_builtin_object_size (tree);
 static rtx expand_builtin_memory_chk (tree, rtx, enum machine_mode,
@@ -10600,6 +10601,9 @@ fold_builtin_3 (location_t loc, tree fnd
     case BUILT_IN_SPRINTF:
       return fold_builtin_sprintf (loc, arg0, arg1, arg2, ignore);
 
+    case BUILT_IN_SNPRINTF:
+      return fold_builtin_snprintf (loc, arg0, arg1, arg2, NULL_TREE, ignore);
+
     case BUILT_IN_STRCPY_CHK:
     case BUILT_IN_STPCPY_CHK:
       return fold_builtin_stxcpy_chk (loc, fndecl, arg0, arg1, arg2, NULL_TREE,
@@ -10665,6 +10669,9 @@ fold_builtin_4 (location_t loc, tree fnd
     case BUILT_IN_STRNCAT_CHK:
       return fold_builtin_strncat_chk (loc, fndecl, arg0, arg1, arg2, arg3);
 
+    case BUILT_IN_SNPRINTF:
+      return fold_builtin_snprintf (loc, arg0, arg1, arg2, arg3, ignore);
+
     case BUILT_IN_FPRINTF_CHK:
     case BUILT_IN_VFPRINTF_CHK:
       if (!validate_arg (arg1, INTEGER_TYPE)
@@ -11923,6 +11930,127 @@ fold_builtin_sprintf (location_t loc, tr
     return call;
 }
 
+/* Simplify a call to the snprintf builtin with arguments DEST, DESTSIZE,
+   FMT, and ORIG.  ORIG may be null if this is a 3-argument call.  We don't
+   attempt to simplify calls with more than 4 arguments.
+
+   Return NULL_TREE if no simplification was possible, otherwise return the
+   simplified form of the call as a tree.  If IGNORED is true, it means that
+   the caller does not use the returned value of the function.  */
+
+static tree
+fold_builtin_snprintf (location_t loc, tree dest, tree destsize, tree fmt,
+		       tree orig, int ignored)
+{
+  tree call, retval;
+  const char *fmt_str = NULL;
+  unsigned HOST_WIDE_INT destlen;
+
+  /* Verify the required arguments in the original call.  We deal with two
+     types of snprintf() calls: 'snprintf (str, cst, fmt)' and
+     'snprintf (dest, cst, "%s", orig)'.  */
+  if (!validate_arg (dest, POINTER_TYPE)
+      || !validate_arg (destsize, INTEGER_TYPE)
+      || !validate_arg (fmt, POINTER_TYPE))
+    return NULL_TREE;
+  if (orig && !validate_arg (orig, POINTER_TYPE))
+    return NULL_TREE;
+
+  if (!host_integerp (destsize, 1))
+    return NULL_TREE;
+
+  /* Check whether the format is a literal string constant.  */
+  fmt_str = c_getstr (fmt);
+  if (fmt_str == NULL)
+    return NULL_TREE;
+
+  call = NULL_TREE;
+  retval = NULL_TREE;
+
+  if (!init_target_chars ())
+    return NULL_TREE;
+
+  destlen = tree_low_cst (destsize, 1);
+
+  /* If the format doesn't contain % args or %%, use strcpy.  */
+  if (strchr (fmt_str, target_percent) == NULL)
+    {
+      tree fn = implicit_built_in_decls[BUILT_IN_STRCPY];
+      size_t len = strlen (fmt_str);
+
+      /* Don't optimize snprintf (buf, 4, "abc", ptr++).  */
+      if (orig)
+	return NULL_TREE;
+
+      /* We could expand this as
+	 memcpy (str, fmt, cst - 1); str[cst - 1] = '\0';
+	 or to
+	 memcpy (str, fmt_with_nul_at_cstm1, cst);
+	 but in the former case that might increase code size
+	 and in the latter case grow .rodata section too much.
+	 So punt for now.  */
+      if (len >= destlen)
+	return NULL_TREE;
+
+      if (!fn)
+	return NULL_TREE;
+
+      /* Convert snprintf (str, cst, fmt) into strcpy (str, fmt) when
+	 'format' is known to contain no % formats and
+	 strlen (fmt) < cst.  */
+      call = build_call_expr_loc (loc, fn, 2, dest, fmt);
+
+      if (!ignored)
+	retval = build_int_cst (NULL_TREE, strlen (fmt_str));
+    }
+
+  /* If the format is "%s", use strcpy if the result isn't used.  */
+  else if (fmt_str && strcmp (fmt_str, target_percent_s) == 0)
+    {
+      tree fn = implicit_built_in_decls[BUILT_IN_STRCPY];
+      unsigned HOST_WIDE_INT origlen;
+
+      /* Don't crash on snprintf (str1, cst, "%s").  */
+      if (!orig)
+	return NULL_TREE;
+
+      retval = c_strlen (orig, 1);
+      if (!retval || !host_integerp (retval, 1))  
+	return NULL_TREE;
+
+      origlen = tree_low_cst (retval, 1);
+      /* We could expand this as
+	 memcpy (str1, str2, cst - 1); str1[cst - 1] = '\0';
+	 or to
+	 memcpy (str1, str2_with_nul_at_cstm1, cst);
+	 but in the former case that might increase code size
+	 and in the latter case grow .rodata section too much.
+	 So punt for now.  */
+      if (origlen >= destlen)
+	return NULL_TREE;
+
+      /* Convert snprintf (str1, cst, "%s", str2) into
+	 strcpy (str1, str2) if strlen (str2) < cst.  */
+      if (!fn)
+	return NULL_TREE;
+
+      call = build_call_expr_loc (loc, fn, 2, dest, orig);
+
+      if (ignored)
+	retval = NULL_TREE;
+    }
+
+  if (call && retval)
+    {
+      retval = fold_convert_loc
+	(loc, TREE_TYPE (TREE_TYPE (implicit_built_in_decls[BUILT_IN_SNPRINTF])),
+	 retval);
+      return build2 (COMPOUND_EXPR, TREE_TYPE (retval), call, retval);
+    }
+  else
+    return call;
+}
+
 /* Expand a call EXP to __builtin_object_size.  */
 
 rtx
--- gcc/testsuite/gcc.c-torture/execute/pr47917.c.jj	2011-02-28 14:04:39.000000000 +0100
+++ gcc/testsuite/gcc.c-torture/execute/pr47917.c	2011-02-28 14:01:05.000000000 +0100
@@ -0,0 +1,32 @@
+/* PR middle-end/47917 */
+
+extern int snprintf (char *, __SIZE_TYPE__, const char *, ...);
+extern int memcmp (const void *, const void *, __SIZE_TYPE__);
+extern void abort (void);
+
+char buf1[6], buf2[6], buf3[4], buf4[4];
+int i;
+
+int
+foo (void)
+{
+  int ret = snprintf (buf1, sizeof buf1, "abcde");
+  ret += snprintf (buf2, sizeof buf2, "abcdef") * 16;
+  ret += snprintf (buf3, sizeof buf3, "%s", i++ < 6 ? "abc" : "def") * 256;
+  ret += snprintf (buf4, sizeof buf4, "%s", i++ > 10 ? "abcde" : "defgh") * 4096;
+  return ret;
+}
+
+int
+main (void)
+{
+  if (foo () != 5 + 6 * 16 + 3 * 256 + 5 * 4096)
+    abort ();
+  if (memcmp (buf1, "abcde", 6) != 0
+      || memcmp (buf2, "abcde", 6) != 0
+      || memcmp (buf3, "abc", 4) != 0
+      || memcmp (buf4, "def", 4) != 0
+      || i != 2)
+    abort ();
+  return 0;
+}
--- gcc/testsuite/gcc.dg/pr47917.c.jj	2011-02-28 14:04:58.000000000 +0100
+++ gcc/testsuite/gcc.dg/pr47917.c	2011-02-28 14:07:22.000000000 +0100
@@ -0,0 +1,38 @@
+/* PR middle-end/47917 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+
+extern int snprintf (char *, __SIZE_TYPE__, const char *, ...);
+extern int memcmp (const void *, const void *, __SIZE_TYPE__);
+extern void abort (void);
+
+char buf1[6], buf2[6], buf3[4], buf4[4];
+int i;
+
+int
+foo (void)
+{
+  int ret = snprintf (buf1, sizeof buf1, "abcde");
+  ret += snprintf (buf2, sizeof buf2, "abcdef") * 16;
+  ret += snprintf (buf3, sizeof buf3, "%s", i++ < 6 ? "abc" : "def") * 256;
+  ret += snprintf (buf4, sizeof buf4, "%s", i++ > 10 ? "abcde" : "defgh") * 4096;
+  return ret;
+}
+
+int
+main (void)
+{
+  if (foo () != 5 + 6 * 16 + 3 * 256 + 5 * 4096)
+    abort ();
+  if (memcmp (buf1, "abcde", 6) != 0
+      || memcmp (buf2, "abcde", 6) != 0
+      || memcmp (buf3, "abc", 4) != 0
+      || memcmp (buf4, "def", 4) != 0
+      || i != 2)
+    abort ();
+  return 0;
+}
+
+/* { dg-final { scan-tree-dump-times "snprintf" 2 "optimized" } } */
+/* { dg-final { scan-tree-dump-times "sprintf" 0 "optimized" } } */
+/* { dg-final { cleanup-tree-dump "optimized" } } */

	Jakub