i386.md (movstrsi, clrstrsi): Support variable sized copies, align destination when needed.

	* i386.md (movstrsi, clrstrsi): Support variable sized copies, align
	destination when needed.
	(strmovsi, strsetsi): New expander.
	(strmovsi_1, strsetsi_1): New pattern.
	* i386.h (MASK_NO_ALIGN_STROP, MASK_INLINE_ALL_STROP,
	TARGET_ALIGN_STRINGOPS, TARGET_INLINE_ALL_STRINGOPS): New macros.
	(TARGET_SWITCHES) Add align-stringops and inline-all-stringops.
	* invoke.texi (align-stringops, inline-all-stringops): Document.

From-SVN: r31773

gcc/ChangeLog

+Thu Feb  3 15:08:13 MET 2000  Jan Hubicka  <jh@suse.cz>
+
+	* i386.md (movstrsi, clrstrsi): Support variable sized copies, align
+	destination when needed.
+	(strmovsi, strsetsi): New expander.
+	(strmovsi_1, strsetsi_1): New pattern.
+	* i386.h (MASK_NO_ALIGN_STROP, MASK_INLINE_ALL_STROP,
+	TARGET_ALIGN_STRINGOPS, TARGET_INLINE_ALL_STRINGOPS): New macros.
+	(TARGET_SWITCHES) Add align-stringops and inline-all-stringops.
+	* invoke.texi (align-stringops, inline-all-stringops): Document.
+
 Wed Feb  2 23:04:47 2000   Krister Walfridsson <cato@df.lth.se>
 
 	* i386/netbsd.h (INT_ASM_OP): Define.

gcc/config/i386/i386.h

@@ -101,6 +101,8 @@ extern int target_flags;
 #define MASK_NO_FANCY_MATH_387	0x00000040	/* Disable sin, cos, sqrt */
 #define MASK_OMIT_LEAF_FRAME_POINTER 0x080      /* omit leaf frame pointers */
 #define MASK_STACK_PROBE	0x00000100	/* Enable stack probing */
+#define MASK_NO_ALIGN_STROPS	0x00001000	/* Enable aligning of string ops. */
+#define MASK_INLINE_ALL_STROPS	0x00002000	/* Inline stringops in all cases */
 
 /* Temporary codegen switches */
 #define MASK_INTEL_SYNTAX	0x00000200
@@ -190,6 +192,9 @@ extern const int x86_promote_QImode, x86_single_stringop;
 
 #define TARGET_STACK_PROBE (target_flags & MASK_STACK_PROBE)
 
+#define TARGET_ALIGN_STRINGOPS (!(target_flags & MASK_NO_ALIGN_STROPS))
+#define TARGET_INLINE_ALL_STRINGOPS (target_flags & MASK_INLINE_ALL_STROPS)
+
 #define ASSEMBLER_DIALECT ((target_flags & MASK_INTEL_SYNTAX) != 0)
 
 #define TARGET_SWITCHES							      \
@@ -238,6 +243,14 @@ extern const int x86_promote_QImode, x86_single_stringop;
   { "intel-syntax",		MASK_INTEL_SYNTAX,			      \
     "Emit Intel syntax assembler opcodes" },				      \
   { "no-intel-syntax",		-MASK_INTEL_SYNTAX, "" },		      \
+  { "align-stringops",		-MASK_NO_ALIGN_STROPS,			      \
+    "Align destination of the string operations" },			      \
+  { "no-align-stringops",	 MASK_NO_ALIGN_STROPS,			      \
+    "Do not align destination of the string operations" },		      \
+  { "inline-all-strinops",	 MASK_INLINE_ALL_STROPS,		      \
+    "Inline all known string operations" },				      \
+  { "no-inline-all-stringops",	-MASK_INLINE_ALL_STROPS,		      \
+    "Do not inline all known string operations" },			      \
   SUBTARGET_SWITCHES							      \
   { "", TARGET_DEFAULT, 0 }}
 

gcc/invoke.texi

@@ -360,7 +360,7 @@ in the following sections.
 -mreg-alloc=@var{list}  -mregparm=@var{num}
 -malign-jumps=@var{num}  -malign-loops=@var{num}
 -malign-functions=@var{num} -mpreferred-stack-boundary=@var{num}
--mthreads
+-mthreads -mno-align-stringops -minline-all-stringops
 
 @emph{HPPA Options}
 -march=@var{architecture type}
@@ -5954,6 +5954,19 @@ on thread-safe exception handling must compile and link all code with the
 @samp{-mthreads} option. When compiling, @samp{-mthreads} defines 
 @samp{-D_MT}; when linking, it links in a special thread helper library 
 @samp{-lmingwthrd} which cleans up per thread exception handling data.
+
+@item -mno-align-stringops
+@kindex -mno-align-stringops
+Do not align destination of inlined string operations. This switch reduces
+code size and improves performance in case the destination is already aligned,
+but gcc don't know about it.
+
+@item -minline-all-stringops
+@kindex -minline-all-stringops
+By default GCC inlines string operations only when destination is known to be
+aligned at least to 4 byte boundary. This enables more inlining, increase code
+size, but may improve performance of code that depends on fast memcpy, strlen
+and memset for short lengths.
 @end table
 
 @node HPPA Options