acconfig.h (HAVE_GAS_MAX_SKIP_P2ALIGN): New tag.

	* acconfig.h (HAVE_GAS_MAX_SKIP_P2ALIGN): New tag.
	* configure.in: Check for it.
	* i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Use it.
	* final.c (uid_align, uid_shuid, label_align): Make static.
	(label_align): Change type to struct label_alignment pointer.
	(LABEL_TO_ALIGNMENT, shorten_branches): Update due to type change.
	(LABEL_TO_MAX_SKIP): Define.
	(LABEL_ALIGN_MAX_SKIP, LOOP_ALIGN_MAX_SKIP,
	LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP): Provide defaults.
	(shorten_branches): Record the maximum bytes to skip when
	aligning a label.
	(final_scan_insn): Use the maximum bytes to skip when aligning a label
	if ASM_OUTPUT_MAX_SKIP_ALIGN is available.
	* i386.h (LOOP_ALIGN_MAX_SKIP,
	LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP): Define.
	* i386.c (override_options): i386_align_jumps and i386_align_loops
	default to 4 if ASM_OUTPUT_MAX_SKIP_ALIGN is available.
	* invoke.texi: Document new i386 align-loops and align-jumps behavior.

From-SVN: r19933

gcc/ChangeLog

+Wed May 20 23:44:28 EDT 1998  John Wehle  (john@feith.com)
+
+	* acconfig.h (HAVE_GAS_MAX_SKIP_P2ALIGN): New tag.
+	* configure.in: Check for it.
+	* i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Use it.
+	* final.c (uid_align, uid_shuid, label_align): Make static.
+	(label_align): Change type to struct label_alignment pointer.
+	(LABEL_TO_ALIGNMENT, shorten_branches): Update due to type change.
+	(LABEL_TO_MAX_SKIP): Define.
+	(LABEL_ALIGN_MAX_SKIP, LOOP_ALIGN_MAX_SKIP,
+	LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP): Provide defaults.
+	(shorten_branches): Record the maximum bytes to skip when
+	aligning a label.
+	(final_scan_insn): Use the maximum bytes to skip when aligning a label
+	if ASM_OUTPUT_MAX_SKIP_ALIGN is available.
+	* i386.h (LOOP_ALIGN_MAX_SKIP,
+	LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP): Define.
+	* i386.c (override_options): i386_align_jumps and i386_align_loops
+	default to 4 if ASM_OUTPUT_MAX_SKIP_ALIGN is available.
+	* invoke.texi: Document new i386 align-loops and align-jumps behavior.
+
 1998-05-21  Mark Mitchell  <mmitchell@usa.net>
 
 	* cplus-dem.c (do_type): Handle volatile qualification.

gcc/acconfig.h

@@ -10,6 +10,10 @@
 /* Define if your compiler understands volatile.  */
 #undef HAVE_VOLATILE
 
+/* Define if your assembler supports specifying the maximum number
+   of bytes to skip when using the GAS .p2align command. */
+#undef HAVE_GAS_MAX_SKIP_P2ALIGN
+
 /* Define if you have a working <inttypes.h> header file.  */
 #undef HAVE_INTTYPES_H
 

gcc/config.in

@@ -11,6 +11,10 @@
 /* Define if your compiler understands volatile.  */
 #undef HAVE_VOLATILE
 
+/* Define if your assembler supports specifying the maximum number
+   of bytes to skip when using the GAS .p2align command. */
+#undef HAVE_GAS_MAX_SKIP_P2ALIGN
+
 /* Define if you have a working <inttypes.h> header file.  */
 #undef HAVE_INTTYPES_H
 

gcc/config/i386/gas.h

@@ -85,6 +85,19 @@ Boston, MA 02111-1307, USA.  */
 #define ASM_OUTPUT_ALIGN(FILE,LOG) \
   if ((LOG)!=0) fprintf ((FILE), "\t.balign %d\n", 1<<(LOG))
 #endif
+
+/* A C statement to output to the stdio stream FILE an assembler
+   command to advance the location counter to a multiple of 1<<LOG
+   bytes if it is within MAX_SKIP bytes.
+
+   This is used to align code labels according to Intel recommendations.  */
+
+#ifdef HAVE_GAS_MAX_SKIP_P2ALIGN
+#  define ASM_OUTPUT_MAX_SKIP_ALIGN(FILE,LOG,MAX_SKIP) \
+     if ((LOG)!=0) \
+       if ((MAX_SKIP)==0) fprintf ((FILE), "\t.p2align %d\n", (LOG)); \
+       else fprintf ((FILE), "\t.p2align %d,,%d\n", (LOG), (MAX_SKIP))
+#endif
 
 /* A C statement or statements which output an assembler instruction
    opcode to the stdio stream STREAM.  The macro-operand PTR is a

gcc/config/i386/i386.c

@@ -332,7 +332,11 @@ override_options ()
 	       i386_align_loops, MAX_CODE_ALIGN);
     }
   else
+#ifdef ASM_OUTPUT_MAX_SKIP_ALIGN
+    i386_align_loops = 4;
+#else
     i386_align_loops = 2;
+#endif
 
   /* Validate -malign-jumps= value, or provide default.  */
   if (i386_align_jumps_string)
@@ -343,7 +347,11 @@ override_options ()
 	       i386_align_jumps, MAX_CODE_ALIGN);
     }
   else
+#ifdef ASM_OUTPUT_MAX_SKIP_ALIGN
+    i386_align_jumps = 4;
+#else
     i386_align_jumps = def_align;
+#endif
 
   /* Validate -malign-functions= value, or provide default. */
   if (i386_align_funcs_string)

gcc/config/i386/i386.h

@@ -506,11 +506,13 @@ extern int ix86_arch;
 
 /* Align loop starts for optimal branching.  */
 #define LOOP_ALIGN(LABEL) (i386_align_loops)
+#define LOOP_ALIGN_MAX_SKIP (i386_align_loops_string ? 0 : 7)
 
 /* This is how to align an instruction for optimal branching.
    On i486 we'll get better performance by aligning on a
    cache line (i.e. 16 byte) boundary.  */
 #define LABEL_ALIGN_AFTER_BARRIER(LABEL) (i386_align_jumps)
+#define LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (i386_align_jumps_string ? 0 : 7)
 
 
 /* Standard register usage.  */

gcc/configure

@@ -5070,6 +5070,53 @@ if [ -f ../ld/Makefile ]; then
 #	fi
 fi
 
+# Figure out what assembler alignment features are present.
+echo $ac_n "checking assembler alignment features""... $ac_c" 1>&6
+echo "configure:5076: checking assembler alignment features" >&5
+gcc_cv_as=
+gcc_cv_as_alignment_features=
+if [ -x as$host_exeext ]; then
+	# Build using assembler in the current directory.
+	gcc_cv_as=./as$host_exeext
+elif [ -f $srcdir/../gas/configure.in ]; then
+	# Single tree build which includes gas.
+	for f in $srcdir/../gas/configure.in $srcdir/../gas/Makefile.in
+	do
+		gcc_cv_gas_version=`grep '^VERSION=[0-9]*\.[0-9]*' $f`
+		if [ x$gcc_cv_gas_version != x ]; then
+			break
+		fi
+	done
+	gcc_cv_gas_major_version=`expr "$gcc_cv_gas_version" : "VERSION=\([0-9]*\)"`
+	gcc_cv_gas_minor_version=`expr "$gcc_cv_gas_version" : "VERSION=[0-9]*\.\([0-9]*\)"`
+	# Gas version 2.8 and later support specifying the maximum
+	# bytes to skip when using .p2align.
+	if [ "$gcc_cv_gas_major_version" -eq 2 -a "$gcc_cv_gas_minor_version" -ge 8 -o "$gcc_cv_gas_major_version" -gt 2 ]; then
+		gcc_cv_as_alignment_features=".p2align including maximum skip"
+		cat >> confdefs.h <<\EOF
+#define HAVE_GAS_MAX_SKIP_P2ALIGN 1
+EOF
+
+	fi
+elif [ x$host = x$target ]; then
+	# Native build.
+	gcc_cv_as=as$host_exeext
+fi
+if [ x$gcc_cv_as != x ]; then
+	# Check if specifying the maximum bytes to skip when
+	# using .p2align is supported.
+	echo ".p2align 4,,7" > conftest.s
+	if $gcc_cv_as -o conftest.o conftest.s > /dev/null 2>&1; then
+		gcc_cv_as_alignment_features=".p2align including maximum skip"
+		cat >> confdefs.h <<\EOF
+#define HAVE_GAS_MAX_SKIP_P2ALIGN 1
+EOF
+
+	fi
+	rm -f conftest.s conftest.o
+fi
+echo "$ac_t""$gcc_cv_as_alignment_features" 1>&6
+
 # Figure out what language subdirectories are present.
 subdirs=
 for lang in ${srcdir}/*/config-lang.in ..

gcc/configure.in

@@ -3226,6 +3226,46 @@ if [[ -f ../ld/Makefile ]]; then
 #	fi
 fi
 
+# Figure out what assembler alignment features are present.
+AC_MSG_CHECKING(assembler alignment features)
+gcc_cv_as=
+gcc_cv_as_alignment_features=
+if [[ -x as$host_exeext ]]; then
+	# Build using assembler in the current directory.
+	gcc_cv_as=./as$host_exeext
+elif [[ -f $srcdir/../gas/configure.in ]]; then
+	# Single tree build which includes gas.
+	for f in $srcdir/../gas/configure.in $srcdir/../gas/Makefile.in
+	do
+		gcc_cv_gas_version=`grep '^VERSION=[[0-9]]*\.[[0-9]]*' $f`
+		if [[ x$gcc_cv_gas_version != x ]]; then
+			break
+		fi
+	done
+	gcc_cv_gas_major_version=`expr "$gcc_cv_gas_version" : "VERSION=\([[0-9]]*\)"`
+	gcc_cv_gas_minor_version=`expr "$gcc_cv_gas_version" : "VERSION=[[0-9]]*\.\([[0-9]]*\)"`
+	# Gas version 2.8 and later support specifying the maximum
+	# bytes to skip when using .p2align.
+	if [[ "$gcc_cv_gas_major_version" -eq 2 -a "$gcc_cv_gas_minor_version" -ge 8 -o "$gcc_cv_gas_major_version" -gt 2 ]]; then
+		gcc_cv_as_alignment_features=".p2align including maximum skip"
+		AC_DEFINE(HAVE_GAS_MAX_SKIP_P2ALIGN)
+	fi
+elif [[ x$host = x$target ]]; then
+	# Native build.
+	gcc_cv_as=as$host_exeext
+fi
+if [[ x$gcc_cv_as != x ]]; then
+	# Check if specifying the maximum bytes to skip when
+	# using .p2align is supported.
+	echo ".p2align 4,,7" > conftest.s
+	if $gcc_cv_as -o conftest.o conftest.s > /dev/null 2>&1; then
+		gcc_cv_as_alignment_features=".p2align including maximum skip"
+		AC_DEFINE(HAVE_GAS_MAX_SKIP_P2ALIGN)
+	fi
+	rm -f conftest.s conftest.o
+fi
+AC_MSG_RESULT($gcc_cv_as_alignment_features)
+
 # Figure out what language subdirectories are present.
 subdirs=
 for lang in ${srcdir}/*/config-lang.in ..

gcc/final.c

@@ -649,9 +649,14 @@ int insn_current_align;
    for each insn we'll call the alignment chain of this insn in the following
    comments.  */
 
-rtx *uid_align;
-int *uid_shuid;
-short *label_align;
+struct label_alignment {
+  short alignment;
+  short max_skip;
+};
+
+static rtx *uid_align;
+static int *uid_shuid;
+static struct label_alignment *label_align;
 
 /* Indicate that branch shortening hasn't yet been done.  */
 
@@ -794,14 +799,26 @@ get_attr_length (insn)
 #define LABEL_ALIGN(LABEL) 0
 #endif
 
+#ifndef LABEL_ALIGN_MAX_SKIP
+#define LABEL_ALIGN_MAX_SKIP 0
+#endif
+
 #ifndef LOOP_ALIGN
 #define LOOP_ALIGN(LABEL) 0
 #endif
 
+#ifndef LOOP_ALIGN_MAX_SKIP
+#define LOOP_ALIGN_MAX_SKIP 0
+#endif
+
 #ifndef LABEL_ALIGN_AFTER_BARRIER
 #define LABEL_ALIGN_AFTER_BARRIER(LABEL) 0
 #endif
 
+#ifndef LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
+#define LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP 0
+#endif
+
 #ifndef ADDR_VEC_ALIGN
 int
 final_addr_vec_align (addr_vec)
@@ -826,7 +843,10 @@ final_addr_vec_align (addr_vec)
 static int min_labelno, max_labelno;
 
 #define LABEL_TO_ALIGNMENT(LABEL) \
-  (label_align[CODE_LABEL_NUMBER (LABEL) - min_labelno])
+  (label_align[CODE_LABEL_NUMBER (LABEL) - min_labelno].alignment)
+
+#define LABEL_TO_MAX_SKIP(LABEL) \
+  (label_align[CODE_LABEL_NUMBER (LABEL) - min_labelno].max_skip)
 
 /* For the benefit of port specific code do this also as a function.  */
 int
@@ -962,6 +982,7 @@ shorten_branches (first)
   int max_uid;
   int i;
   int max_log;
+  int max_skip;
 #ifdef HAVE_ATTR_length
 #define MAX_CODE_ALIGN 16
   rtx seq;
@@ -1001,10 +1022,10 @@ shorten_branches (first)
 
   max_labelno = max_label_num ();
   min_labelno = get_first_label_num ();
-  label_align
-    = (short*) xmalloc ((max_labelno - min_labelno + 1) * sizeof (short));
-  bzero ((char *) label_align,
-	 (max_labelno - min_labelno + 1) * sizeof (short));
+  label_align = (struct label_alignment *) xmalloc (
+    (max_labelno - min_labelno + 1) * sizeof (struct label_alignment));
+  bzero (label_align,
+    (max_labelno - min_labelno + 1) * sizeof (struct label_alignment));
 
   uid_shuid = (int *) xmalloc (max_uid * sizeof *uid_shuid);
 
@@ -1014,7 +1035,10 @@ shorten_branches (first)
      impose on the next CODE_LABEL (or the current one if we are processing
      the CODE_LABEL itself).  */
      
-  for (max_log = 0, insn = get_insns (), i = 1; insn; insn = NEXT_INSN (insn))
+  max_log = 0;
+  max_skip = 0;
+
+  for (insn = get_insns (), i = 1; insn; insn = NEXT_INSN (insn))
     {
       int log;
 
@@ -1033,7 +1057,10 @@ shorten_branches (first)
 
 	  log = LABEL_ALIGN (insn);
 	  if (max_log < log)
-	    max_log = log;
+	    {
+	      max_log = log;
+	      max_skip = LABEL_ALIGN_MAX_SKIP;
+	    }
 	  next = NEXT_INSN (insn);
 /* ADDR_VECs only take room if read-only data goes into the text section.  */
 #if !defined(READONLY_DATA_SECTION) || defined(JUMP_TABLES_IN_TEXT_SECTION)
@@ -1045,12 +1072,17 @@ shorten_branches (first)
 		{
 		  log = ADDR_VEC_ALIGN (next);
 		  if (max_log < log)
-		    max_log = log;
+		    {
+		      max_log = log;
+		      max_skip = LABEL_ALIGN_MAX_SKIP;
+		    }
 		}
 	    }
 #endif
 	  LABEL_TO_ALIGNMENT (insn) = max_log;
+	  LABEL_TO_MAX_SKIP (insn) = max_skip;
 	  max_log = 0;
+	  max_skip = 0;
 	}
       else if (GET_CODE (insn) == BARRIER)
 	{
@@ -1062,7 +1094,10 @@ shorten_branches (first)
 	      {
 		log = LABEL_ALIGN_AFTER_BARRIER (insn);
 		if (max_log < log)
-		  max_log = log;
+		  {
+		    max_log = log;
+		    max_skip = LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP;
+		  }
 		break;
 	      }
 	}
@@ -1078,7 +1113,10 @@ shorten_branches (first)
 	      {
 		log = LOOP_ALIGN (insn);
 		if (max_log < log)
-		  max_log = log;
+		  {
+		    max_log = log;
+		    max_skip = LOOP_ALIGN_MAX_SKIP;
+		  }
 		break;
 	      }
 	}
@@ -2222,9 +2260,14 @@ final_scan_insn (insn, file, optimize, prescan, nopeepholes)
       if (CODE_LABEL_NUMBER (insn) <= max_labelno)
 	{
 	  int align = LABEL_TO_ALIGNMENT (insn);
+	  int max_skip = LABEL_TO_MAX_SKIP (insn);
 
 	  if (align && NEXT_INSN (insn))
+#ifdef ASM_OUTPUT_MAX_SKIP_ALIGN
+	    ASM_OUTPUT_MAX_SKIP_ALIGN (file, align, max_skip);
+#else
 	    ASM_OUTPUT_ALIGN (file, align);
+#endif
 	}
       CC_STATUS_INIT;
       if (prescan > 0)

gcc/invoke.texi

@@ -4905,12 +4905,18 @@ startup modules.
 
 @item -malign-loops=@var{num}
 Align loops to a 2 raised to a @var{num} byte boundary.  If
-@samp{-malign-loops} is not specified, the default is 2.
+@samp{-malign-loops} is not specified, the default is 2 unless
+gas 2.8 (or later) is being used in which case the default is
+to align the loop on a 16 byte boundary if it is less than 8
+bytes away.
 
 @item -malign-jumps=@var{num}
 Align instructions that are only jumped to to a 2 raised to a @var{num}
 byte boundary.  If @samp{-malign-jumps} is not specified, the default is
-2 if optimizing for a 386, and 4 if optimizing for a 486.
+2 if optimizing for a 386, and 4 if optimizing for a 486 unless
+gas 2.8 (or later) is being used in which case the default is
+to align the instruction on a 16 byte boundary if it is less
+than 8 bytes away.
 
 @item -malign-functions=@var{num}
 Align the start of functions to a 2 raised to @var{num} byte boundary.