x86: Remove the prefix byte from non-VEX/EVEX base_opcode

Replace the prefix byte in non-VEX/EVEX base_opcode with PREFIX_0X66,
PREFIX_0XF2 or PREFIX_0XF3.

gas/

	* config/tc-i386.c (load_insn_p): Check opcodeprefix == 0 for
	base_opcode == 0xfc7.
	(match_template): Likewise.
	(process_suffix): Check opcodeprefix == PREFIX_0XF2 for CRC32.
	(check_byte_reg): Likewise.
	(output_insn): Don't add the 0xf3 prefix twice for PadLock
	instructions.  Don't add prefix from non-VEX/EVEX base_opcode.

opcodes/

	* i386-gen.c (process_i386_opcode_modifier): Return 1 for
	non-VEX/EVEX/prefix encoding.
	(output_i386_opcode): Fail if non-VEX/EVEX/prefix base_opcode
	has a prefix byte.
	* i386-opc.tbl: Replace the prefix byte in non-VEX/EVEX
	base_opcode with PREFIX_0X66, PREFIX_0XF2 or PREFIX_0XF3.
	* i386-tbl.h: Regenerated.

gas/ChangeLog

@@ -1,3 +1,13 @@
+2020-10-14  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* config/tc-i386.c (load_insn_p): Check opcodeprefix == 0 for
+	base_opcode == 0xfc7.
+	(match_template): Likewise.
+	(process_suffix): Check opcodeprefix == PREFIX_0XF2 for CRC32.
+	(check_byte_reg): Likewise.
+	(output_insn): Don't add the 0xf3 prefix twice for PadLock
+	instructions.  Don't add prefix from non-VEX/EVEX base_opcode.
+
 2020-10-13  H.J. Lu  <hongjiu.lu@intel.com>
 
 	* config/tc-i386.c (build_vex_prefix): Replace vexopcode with

gas/config/tc-i386.c

@@ -4434,6 +4434,7 @@ load_insn_p (void)
 
       /* cmpxchg8b, cmpxchg16b, xrstors.  */
       if (i.tm.base_opcode == 0xfc7
+	  && i.tm.opcode_modifier.opcodeprefix == 0
 	  && (i.tm.extension_opcode == 1 || i.tm.extension_opcode == 3))
 	return 1;
 
@@ -4453,6 +4454,7 @@ load_insn_p (void)
 
       /* vmptrld */
       if (i.tm.base_opcode == 0xfc7
+	  && i.tm.opcode_modifier.opcodeprefix == 0
 	  && i.tm.extension_opcode == 6)
 	return 1;
 
@@ -6313,8 +6315,9 @@ match_template (char mnem_suffix)
       j = i.imm_operands + (t->operands > i.imm_operands + 1);
       if (((i.suffix == QWORD_MNEM_SUFFIX
 	    && flag_code != CODE_64BIT
-	    && (t->base_opcode != 0x0fc7
+	    && !(t->base_opcode == 0xfc7
-		|| t->extension_opcode != 1 /* cmpxchg8b */))
+		 && i.tm.opcode_modifier.opcodeprefix == 0
+		 && t->extension_opcode == 1) /* cmpxchg8b */)
 	   || (i.suffix == LONG_MNEM_SUFFIX
 	       && !cpu_arch_flags.bitfield.cpui386))
 	  && (intel_syntax
@@ -6759,6 +6762,8 @@ check_string (void)
 static int
 process_suffix (void)
 {
+  bfd_boolean is_crc32 = FALSE;
+
   /* If matched instruction specifies an explicit instruction mnemonic
      suffix, use it.  */
   if (i.tm.opcode_modifier.size == SIZE16)
@@ -6772,6 +6777,9 @@ process_suffix (void)
 	   && !i.tm.opcode_modifier.addrprefixopreg)
     {
       unsigned int numop = i.operands;
+      /* CRC32 */
+      is_crc32 = (i.tm.base_opcode == 0xf38f0
+		  && i.tm.opcode_modifier.opcodeprefix == PREFIX_0XF2);
 
       /* movsx/movzx want only their source operand considered here, for the
 	 ambiguity checking below.  The suffix will be replaced afterwards
@@ -6781,8 +6789,7 @@ process_suffix (void)
 	--i.operands;
 
       /* crc32 needs REX.W set regardless of suffix / source operand size.  */
-      if (i.tm.base_opcode == 0xf20f38f0
+      if (is_crc32 && i.tm.operand_types[1].bitfield.qword)
-          && i.tm.operand_types[1].bitfield.qword)
         i.rex |= REX_W;
 
       /* If there's no instruction mnemonic suffix we try to invent one
@@ -6793,7 +6800,7 @@ process_suffix (void)
 	     Destination register type is more significant than source
 	     register type.  crc32 in SSE4.2 prefers source register
 	     type. */
-	  unsigned int op = i.tm.base_opcode != 0xf20f38f0 ? i.operands : 1;
+	  unsigned int op = is_crc32 ? 1 : i.operands;
 
 	  while (op--)
 	    if (i.tm.operand_types[op].bitfield.instance == InstanceNone
@@ -7143,7 +7150,7 @@ process_suffix (void)
 		      || i.tm.operand_types[0].bitfield.instance == RegD
 		      || i.tm.operand_types[1].bitfield.instance == RegD
 		      /* CRC32 */
-		      || i.tm.base_opcode == 0xf20f38f0))))
+		      || is_crc32))))
 	i.tm.base_opcode |= 1;
       break;
     }
@@ -7257,7 +7264,9 @@ check_byte_reg (void)
 	continue;
 
       /* crc32 only wants its source operand checked here.  */
-      if (i.tm.base_opcode == 0xf20f38f0 && op)
+      if (i.tm.base_opcode == 0xf38f0
+	  && i.tm.opcode_modifier.opcodeprefix == PREFIX_0XF2
+	  && op != 0)
 	continue;
 
       /* Any other register is bad.  */
@@ -9302,7 +9311,6 @@ output_insn (void)
       char *p;
       unsigned char *q;
       unsigned int j;
-      unsigned int prefix;
       enum mf_cmp_kind mf_cmp;
 
       if (avoid_fence
@@ -9382,28 +9390,15 @@ output_insn (void)
 	      add_prefix (0xf2);
 	      break;
 	    case PREFIX_0XF3:
-	      add_prefix (0xf3);
+	      if (!i.tm.cpu_flags.bitfield.cpupadlock
+		  || (i.prefix[REP_PREFIX] != 0xf3))
+		add_prefix (0xf3);
 	      break;
 	    case PREFIX_NONE:
 	      switch (i.tm.opcode_length)
 		{
 		case 3:
-		  if (i.tm.base_opcode & 0xff000000)
-		    {
-		      prefix = (i.tm.base_opcode >> 24) & 0xff;
-		      if (!i.tm.cpu_flags.bitfield.cpupadlock
-			  || prefix != REPE_PREFIX_OPCODE
-			  || (i.prefix[REP_PREFIX] != REPE_PREFIX_OPCODE))
-			add_prefix (prefix);
-		    }
-		  break;
 		case 2:
-		  if ((i.tm.base_opcode & 0xff0000) != 0)
-		    {
-		      prefix = (i.tm.base_opcode >> 16) & 0xff;
-		      add_prefix (prefix);
-		    }
-		  break;
 		case 1:
 		  break;
 		case 0:

opcodes/ChangeLog

@@ -1,3 +1,13 @@
+2020-10-14  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* i386-gen.c (process_i386_opcode_modifier): Return 1 for
+	non-VEX/EVEX/prefix encoding.
+	(output_i386_opcode): Fail if non-VEX/EVEX/prefix base_opcode
+	has a prefix byte.
+	* i386-opc.tbl: Replace the prefix byte in non-VEX/EVEX
+	base_opcode with PREFIX_0X66, PREFIX_0XF2 or PREFIX_0XF3.
+	* i386-tbl.h: Regenerated.
+
 2020-10-13  H.J. Lu  <hongjiu.lu@intel.com>
 
 	* i386-gen.c (opcode_modifiers): Replace VexOpcode with

opcodes/i386-gen.c

@@ -1161,11 +1161,12 @@ adjust_broadcast_modifier (char **opnd)
   return bcst_type;
 }
 
-static void
+static int
 process_i386_opcode_modifier (FILE *table, char *mod, char **opnd, int lineno)
 {
   char *str, *next, *last;
   bitfield modifiers [ARRAY_SIZE (opcode_modifiers)];
+  unsigned int regular_encoding = 1;
 
   active_isstring = 0;
 
@@ -1184,9 +1185,22 @@ process_i386_opcode_modifier (FILE *table, char *mod, char **opnd, int lineno)
 	    {
 	      int val = 1;
 	      if (strcasecmp(str, "Broadcast") == 0)
+		{
 		  val = adjust_broadcast_modifier (opnd);
+		  regular_encoding = 0;
+		}
+	      else if (strcasecmp(str, "Vex") == 0
+		       || strncasecmp(str, "Vex=", 4) == 0
+		       || strcasecmp(str, "EVex") == 0
+		       || strncasecmp(str, "EVex=", 5) == 0
+		       || strncasecmp(str, "Disp8MemShift=", 14) == 0
+		       || strncasecmp(str, "Masking=", 8) == 0
+		       || strcasecmp(str, "SAE") == 0
+		       || strcasecmp(str, "IsPrefix") == 0)
+		regular_encoding = 0;
+
 	      set_bitfield (str, modifiers, val, ARRAY_SIZE (modifiers),
-			  lineno);
+			    lineno);
 	      if (strcasecmp(str, "IsString") == 0)
 		active_isstring = 1;
 
@@ -1215,6 +1229,8 @@ process_i386_opcode_modifier (FILE *table, char *mod, char **opnd, int lineno)
 		 filename, lineno);
     }
   output_opcode_modifier (table, modifiers, ARRAY_SIZE (modifiers));
+
+  return regular_encoding;
 }
 
 enum stage {
@@ -1396,7 +1412,40 @@ output_i386_opcode (FILE *table, const char *name, char *str,
 
   process_i386_cpu_flag (table, cpu_flags, 0, ",", "    ", lineno);
 
-  process_i386_opcode_modifier (table, opcode_modifier, operand_types, lineno);
+  if (process_i386_opcode_modifier (table, opcode_modifier,
+				    operand_types, lineno))
+    {
+      char *end;
+      unsigned long int length = strtoul (opcode_length, &end, 0);
+      unsigned long int opcode = strtoul (base_opcode, &end, 0);
+      switch (length)
+	{
+	case 3:
+	  if ((opcode >> 24) != 0)
+	    fail (_("%s: %s: (base_opcode >> 24) != 0: %s\n"),
+		  filename, name, base_opcode);
+	  break;
+	case 2:
+	  if ((opcode >> 16) != 0)
+	    fail (_("%s: %s: (base_opcode >> 16) != 0: %s\n"),
+		  filename, name, base_opcode);
+	  break;
+	case 1:
+	  if ((opcode >> 8) != 0)
+	    fail (_("%s: %s: (base_opcode >> 8) != 0: %s\n"),
+		  filename, name, base_opcode);
+	  break;
+	case 0:
+	  if (opcode != 0)
+	    fail (_("%s: %s: base_opcode != 0: %s\n"),
+		  filename, name, base_opcode);
+	  break;
+	default:
+	  fail (_("%s: %s: invalid opcode length: %s\n"),
+		filename, name, opcode_length);
+	  break;
+	}
+    }
 
   fprintf (table, "    { ");