[PATCH] Fix builtin_str{,n}cmp bug on s390{,x}

[Jakub Jelinek <jakub at redhat dot com>]

Hi!

On Thu, Jul 10, 2003 at 04:06:16PM -0700, Richard Henderson wrote:
> On Thu, Jul 10, 2003 at 08:38:53PM +0200, Ulrich Weigand wrote:
> > ... we do have an instruction to implement strcmp - CLST - which
> > is currently not used at all by gcc ...).
> 
> Oh yeah?  Then renaming the current pattern to to cmpmemsi
> would seem to be the best solution.

Ok, how about this then (mainline edition)?

2003-07-11  Jakub Jelinek  <jakub@redhat.com>

	* optabs.c (prepare_cmp_insn): Try cmpmemM first if it exists,
	then fall back to cmpstrM.
	* builtins.c (expand_builtin_memcmp): Likewise.
	* config/s390/s390-protos.h (s390_expand_cmpstr): Rename to...
	(s390_expand_cmpmem): ... this.
	* config/s390/s390.md (cmpmemdi, cmpmemsi, cmpmem_short_64,
	cmpmem_short_31, cmpmem_long_64, cmpmem_long_31): Renamed
	from cmpstr* patterns.  Rename call to s390_expand_cmpstr
	to s390_expand_cmpmem.
	* config/s390/s390.c (s390_expand_cmpstr): Rename to...
	(s390_expand_cmpstr): ... this.  Rename cmpstr* instructions
	to cmpmem*.
	* config/i370/i370.md (cmpmemsi, cmpmemsi_1): Renamed from
	cmpstr* patterns.
	* doc/md.texi (cmpstrM): Describe as String compare insn, not
	Block compare insn.
	(cmpmemM): Add.

	* gcc.dg/20030711-1.c: New test.

--- gcc/doc/md.texi.jj	2003-07-11 04:27:20.000000000 -0400
+++ gcc/doc/md.texi	2003-07-11 05:19:32.000000000 -0400
@@ -2799,7 +2799,7 @@ The use for multiple @code{clrstr@var{m}
 
 @cindex @code{cmpstr@var{m}} instruction pattern
 @item @samp{cmpstr@var{m}}
-Block compare instruction, with five operands.  Operand 0 is the output;
+String compare instruction, with five operands.  Operand 0 is the output;
 it has mode @var{m}.  The remaining four operands are like the operands
 of @samp{movstr@var{m}}.  The two memory blocks specified are compared
 byte by byte in lexicographic order starting at the beginning of each
@@ -2809,6 +2809,16 @@ that may access an invalid page or segme
 effect of the instruction is to store a value in operand 0 whose sign
 indicates the result of the comparison.
 
+@cindex @code{cmpmem@var{m}} instruction pattern
+@item @samp{cmpmem@var{m}}
+Block compare instruction, with five operands like the operands
+of @samp{cmpstr@var{m}}.  The two memory blocks specified are compared
+byte by byte in lexicographic order starting at the beginning of each
+block.  Unlike @samp{cmpstr@var{m}} the instruction can prefetch
+any bytes in the two memory blocks.  The effect of the instruction is
+to store a value in operand 0 whose sign indicates the result of the
+comparison.
+
 @cindex @code{strlen@var{m}} instruction pattern
 @item @samp{strlen@var{m}}
 Compute the length of a string, with three operands.
--- gcc/optabs.c.jj	2003-07-11 04:26:42.000000000 -0400
+++ gcc/optabs.c	2003-07-11 05:10:28.000000000 -0400
@@ -3579,6 +3579,40 @@ prepare_cmp_insn (rtx *px, rtx *py, enum
 
       if (size == 0)
 	abort ();
+#ifdef HAVE_cmpmemqi
+      if (HAVE_cmpmemqi
+	  && GET_CODE (size) == CONST_INT
+	  && INTVAL (size) < (1 << GET_MODE_BITSIZE (QImode)))
+	{
+	  result_mode = insn_data[(int) CODE_FOR_cmpmemqi].operand[0].mode;
+	  result = gen_reg_rtx (result_mode);
+	  emit_insn (gen_cmpmemqi (result, x, y, size, opalign));
+	}
+      else
+#endif
+#ifdef HAVE_cmpmemhi
+      if (HAVE_cmpmemhi
+	  && GET_CODE (size) == CONST_INT
+	  && INTVAL (size) < (1 << GET_MODE_BITSIZE (HImode)))
+	{
+	  result_mode = insn_data[(int) CODE_FOR_cmpmemhi].operand[0].mode;
+	  result = gen_reg_rtx (result_mode);
+	  emit_insn (gen_cmpmemhi (result, x, y, size, opalign));
+	}
+      else
+#endif
+#ifdef HAVE_cmpmemsi
+      if (HAVE_cmpmemsi)
+	{
+	  result_mode = insn_data[(int) CODE_FOR_cmpmemsi].operand[0].mode;
+	  result = gen_reg_rtx (result_mode);
+	  size = protect_from_queue (size, 0);
+	  emit_insn (gen_cmpmemsi (result, x, y,
+				   convert_to_mode (SImode, size, 1),
+				   opalign));
+	}
+      else
+#endif
 #ifdef HAVE_cmpstrqi
       if (HAVE_cmpstrqi
 	  && GET_CODE (size) == CONST_INT
--- gcc/config/s390/s390-protos.h.jj	2003-07-11 04:27:10.000000000 -0400
+++ gcc/config/s390/s390-protos.h	2003-07-11 05:03:35.000000000 -0400
@@ -68,7 +68,7 @@ extern void emit_symbolic_move PARAMS ((
 extern void s390_load_address PARAMS ((rtx, rtx));
 extern void s390_expand_movstr PARAMS ((rtx, rtx, rtx));
 extern void s390_expand_clrstr PARAMS ((rtx, rtx));
-extern void s390_expand_cmpstr PARAMS ((rtx, rtx, rtx, rtx));
+extern void s390_expand_cmpmem PARAMS ((rtx, rtx, rtx, rtx));
 extern rtx s390_return_addr_rtx PARAMS ((int, rtx));
 
 extern void s390_output_symbolic_const PARAMS ((FILE *, rtx));
--- gcc/config/s390/s390.md.jj	2003-07-11 04:27:11.000000000 -0400
+++ gcc/config/s390/s390.md	2003-07-11 05:04:04.000000000 -0400
@@ -1958,33 +1958,33 @@
    (set_attr "length"  "8")])
 
 ;
-; cmpstrM instruction pattern(s).
+; cmpmemM instruction pattern(s).
 ;
 
-(define_expand "cmpstrdi"
+(define_expand "cmpmemdi"
   [(set (match_operand:DI 0 "register_operand" "")
         (compare:DI (match_operand:BLK 1 "memory_operand" "")
                     (match_operand:BLK 2 "memory_operand" "") ) )
    (use (match_operand:DI 3 "general_operand" ""))
    (use (match_operand:DI 4 "" ""))]
   "TARGET_64BIT"
-  "s390_expand_cmpstr (operands[0], operands[1], 
+  "s390_expand_cmpmem (operands[0], operands[1], 
                        operands[2], operands[3]); DONE;")
 
-(define_expand "cmpstrsi"
+(define_expand "cmpmemsi"
   [(set (match_operand:SI 0 "register_operand" "")
         (compare:SI (match_operand:BLK 1 "memory_operand" "")
                     (match_operand:BLK 2 "memory_operand" "") ) )
    (use (match_operand:SI 3 "general_operand" ""))
    (use (match_operand:SI 4 "" ""))]
   ""
-  "s390_expand_cmpstr (operands[0], operands[1], 
+  "s390_expand_cmpmem (operands[0], operands[1], 
                        operands[2], operands[3]); DONE;")
 
 ; Compare a block that is up to 256 bytes in length.
 ; The block length is taken as (operands[2] % 256) + 1.
 
-(define_insn "cmpstr_short_64"
+(define_insn "cmpmem_short_64"
   [(set (reg:CCS 33)
         (compare:CCS (match_operand:BLK 0 "memory_operand" "=Q,Q")
                      (match_operand:BLK 1 "memory_operand" "Q,Q")))
@@ -2012,7 +2012,7 @@
    (set_attr "atype"   "*,agen")
    (set_attr "length"  "*,14")])
 
-(define_insn "cmpstr_short_31"
+(define_insn "cmpmem_short_31"
   [(set (reg:CCS 33)
         (compare:CCS (match_operand:BLK 0 "memory_operand" "=Q,Q")
                      (match_operand:BLK 1 "memory_operand" "Q,Q")))
@@ -2042,7 +2042,7 @@
 
 ; Compare a block of arbitrary length.
 
-(define_insn "cmpstr_long_64"
+(define_insn "cmpmem_long_64"
   [(clobber (match_operand:TI 0 "register_operand" "=d"))
    (clobber (match_operand:TI 1 "register_operand" "=d"))
    (set (reg:CCS 33)
@@ -2055,7 +2055,7 @@
   [(set_attr "op_type" "RR")
    (set_attr "type"    "vs")])
 
-(define_insn "cmpstr_long_31"
+(define_insn "cmpmem_long_31"
   [(clobber (match_operand:DI 0 "register_operand" "=d"))
    (clobber (match_operand:DI 1 "register_operand" "=d"))
    (set (reg:CCS 33)
--- gcc/config/s390/s390.c.jj	2003-07-11 04:27:10.000000000 -0400
+++ gcc/config/s390/s390.c	2003-07-11 05:04:34.000000000 -0400
@@ -3116,16 +3116,16 @@ s390_expand_clrstr (dst, len)
    and return the result in TARGET.  */
 
 void
-s390_expand_cmpstr (target, op0, op1, len)
+s390_expand_cmpmem (target, op0, op1, len)
      rtx target;
      rtx op0;
      rtx op1;
      rtx len;
 {
   rtx (*gen_short) PARAMS ((rtx, rtx, rtx)) = 
-    TARGET_64BIT ? gen_cmpstr_short_64 : gen_cmpstr_short_31;
+    TARGET_64BIT ? gen_cmpmem_short_64 : gen_cmpmem_short_31;
   rtx (*gen_long) PARAMS ((rtx, rtx, rtx, rtx)) = 
-    TARGET_64BIT ? gen_cmpstr_long_64 : gen_cmpstr_long_31;
+    TARGET_64BIT ? gen_cmpmem_long_64 : gen_cmpmem_long_31;
   rtx (*gen_result) PARAMS ((rtx)) =
     GET_MODE (target) == DImode ? gen_cmpint_di : gen_cmpint_si;
 
--- gcc/config/i370/i370.md.jj	2003-04-23 15:56:31.000000000 -0400
+++ gcc/config/i370/i370.md	2003-07-11 05:05:59.000000000 -0400
@@ -474,10 +474,10 @@ check_label_emit ();
 )
 
 ;
-; cmpstrsi instruction pattern(s).
+; cmpmemsi instruction pattern(s).
 ;
 
-(define_expand "cmpstrsi"
+(define_expand "cmpmemsi"
   [(set (match_operand:SI 0 "general_operand" "")
 	  (compare (match_operand:BLK 1 "general_operand" "")
 		   (match_operand:BLK 2 "general_operand" "")))
@@ -545,7 +545,7 @@ check_label_emit ();
         emit_move_insn (gen_rtx_SUBREG (SImode, reg2, GET_MODE_SIZE (SImode)), len);
 
         /* Compare! */
-        emit_insn (gen_cmpstrsi_1 (result, reg1, reg2));
+        emit_insn (gen_cmpmemsi_1 (result, reg1, reg2));
     }
   DONE;
 }")
@@ -569,7 +569,7 @@ check_label_emit ();
 
 ; Compare a block that is larger than 255 bytes in length.
 
-(define_insn "cmpstrsi_1"
+(define_insn "cmpmemsi_1"
   [(set (match_operand:SI 0 "register_operand" "+d")
         (compare
         (mem:BLK (subreg:SI (match_operand:DI 1 "register_operand" "+d") 0))
--- gcc/builtins.c.jj	2003-07-09 06:26:00.000000000 -0400
+++ gcc/builtins.c	2003-07-11 06:02:12.000000000 -0400
@@ -3196,7 +3196,7 @@ expand_builtin_memcmp (tree exp ATTRIBUT
       return expand_expr (result, target, mode, EXPAND_NORMAL);
     }
 
-#ifdef HAVE_cmpstrsi
+#if defined HAVE_cmpmemsi || defined HAVE_cmpstrsi
   {
     rtx arg1_rtx, arg2_rtx, arg3_rtx;
     rtx result;
@@ -3206,8 +3206,19 @@ expand_builtin_memcmp (tree exp ATTRIBUT
       = get_pointer_alignment (arg1, BIGGEST_ALIGNMENT) / BITS_PER_UNIT;
     int arg2_align
       = get_pointer_alignment (arg2, BIGGEST_ALIGNMENT) / BITS_PER_UNIT;
-    enum machine_mode insn_mode
-      = insn_data[(int) CODE_FOR_cmpstrsi].operand[0].mode;
+    enum machine_mode insn_mode;
+
+#ifdef HAVE_cmpmemsi
+    if (HAVE_cmpmemsi)
+      insn_mode = insn_data[(int) CODE_FOR_cmpmemsi].operand[0].mode;
+    else
+#endif
+#ifdef HAVE_cmpstrsi
+    if (HAVE_cmpstrsi)
+      insn_mode = insn_data[(int) CODE_FOR_cmpstrsi].operand[0].mode;
+    else
+#endif
+      return 0;     
 
     /* If we don't have POINTER_TYPE, call the function.  */
     if (arg1_align == 0 || arg2_align == 0)
@@ -3223,11 +3234,19 @@ expand_builtin_memcmp (tree exp ATTRIBUT
     arg1_rtx = get_memory_rtx (arg1);
     arg2_rtx = get_memory_rtx (arg2);
     arg3_rtx = expand_expr (len, NULL_RTX, VOIDmode, 0);
-    if (!HAVE_cmpstrsi)
-      insn = NULL_RTX;
+#ifdef HAVE_cmpmemsi
+    if (HAVE_cmpmemsi)
+      insn = gen_cmpmemsi (result, arg1_rtx, arg2_rtx, arg3_rtx,
+			   GEN_INT (MIN (arg1_align, arg2_align)));
     else
+#endif
+#ifdef HAVE_cmpstrsi
+    if (HAVE_cmpstrsi)
       insn = gen_cmpstrsi (result, arg1_rtx, arg2_rtx, arg3_rtx,
 			   GEN_INT (MIN (arg1_align, arg2_align)));
+    else
+#endif
+      abort ();
 
     if (insn)
       emit_insn (insn);
--- gcc/testsuite/gcc.dg/20030711-1.c.jj	2003-01-30 05:24:37.000000000 -0500
+++ gcc/testsuite/gcc.dg/20030711-1.c	2003-07-11 06:35:41.000000000 -0400
@@ -0,0 +1,25 @@
+/* Test whether strncmp has not been "optimized" into memcmp
+   nor any code with memcmp semantics.  */
+/* { dg-do run { target i?86-*-linux* x86_64-*-linux* ia64-*-linux* alpha*-*-linux* powerpc*-*-linux* s390*-*-linux* sparc*-*-linux* } } */
+/* { dg-options "-O2" } */
+#include <sys/mman.h>
+#include <stdlib.h>
+
+void __attribute__((noinline)) test (const char *p)
+{
+  if (__builtin_strncmp (p, "abcdefghijklmnopq", 17) == 0)
+    abort ();
+}
+
+int main (void)
+{
+  char *p = mmap (NULL, 131072, PROT_READ | PROT_WRITE,
+                  MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+  if (p == MAP_FAILED)
+    return 0;
+  if (munmap (p + 65536, 65536) < 0)
+    return 0;
+  __builtin_memcpy (p + 65536 - 5, "abcd", 5);
+  test (p + 65536 - 5);
+  return 0;
+}


	Jakub