[RFC] Support syntax and opcodes from the loongson-community/loongarch-opcodes project

gas/config/tc-loongarch.c

@@ -108,6 +108,8 @@ enum options
   OPTION_LA_GLOBAL_WITH_PCREL,
   OPTION_LA_GLOBAL_WITH_ABS,
 
+  OPTION_SYNTAX,
+
   OPTION_END_OF_ENUM,
 };
 
@@ -121,6 +123,8 @@ struct option md_longopts[] =
   { "mla-global-with-pcrel", no_argument, NULL, OPTION_LA_GLOBAL_WITH_PCREL },
   { "mla-global-with-abs", no_argument, NULL, OPTION_LA_GLOBAL_WITH_ABS },
 
+  { "msyntax", required_argument, NULL, OPTION_SYNTAX },
+
   { NULL, no_argument, NULL, 0 }
 };
 
@@ -155,6 +159,14 @@ md_parse_option (int c, const char *arg)
     case OPTION_LA_GLOBAL_WITH_ABS:
       LARCH_opts.la_global_with_abs = 1;
       break;
+    case OPTION_SYNTAX:
+      if (strcasecmp (arg, "vendor") == 0)
+	LARCH_opts.use_community_syntax = 0;
+      else if (strcasecmp (arg, "community") == 0)
+	LARCH_opts.use_community_syntax = 1;
+      else
+	as_fatal (_("invalid -msyntax= option: `%s'"), arg);
+      break;
     case OPTION_IGNORE:
       break;
     }
@@ -338,14 +350,22 @@ s_dtprel (int bytes)
   demand_empty_rest_of_line ();
 }
 
-static const pseudo_typeS loongarch_pseudo_table[] = 
+static void
+s_community_syntax (int enabled)
+{
+	LARCH_opts.use_community_syntax = enabled;
+}
+
+static const pseudo_typeS loongarch_pseudo_table[] =
 {
   { "align", s_loongarch_align, -4 },
   { "dword", cons, 8 },
   { "word", cons, 4 },
   { "half", cons, 2 },
   { "dtprelword", s_dtprel, 4 },
   { "dtpreldword", s_dtprel, 8 },
+  { "community_syntax", s_community_syntax, 1 },
+  { "vendor_syntax", s_community_syntax, 0 },
   { NULL, NULL, 0 },
 };
 
@@ -620,8 +640,22 @@ get_loongarch_opcode (struct loongarch_cl_insn *insn)
             str_hash_insert (ase->name_hash_entry, it->name, (void *) it, 0);
         }
 
+      if (LARCH_opts.use_community_syntax && !ase->community_name_hash_entry)
+        {
+          ase->community_name_hash_entry = str_htab_create ();
+          for (it = ase->opcodes; it->name; it++)
+            if (it->community_name)
+              str_hash_insert (ase->community_name_hash_entry,
+                               it->community_name, (void *) it, 0);
+        }
+
       if ((it = str_hash_find (ase->name_hash_entry, insn->name)) == NULL)
-        continue;
+        {
+          if (!LARCH_opts.use_community_syntax)
+            continue;
+          if ((it = str_hash_find (ase->community_name_hash_entry, insn->name)) == NULL)
+            continue;
+        }
 
       do
         {
@@ -631,7 +665,8 @@ get_loongarch_opcode (struct loongarch_cl_insn *insn)
           insn->arg_num = 0;
           insn->reloc_num = 0;
           insn->insn_bin = loongarch_foreach_args (
-            it->format, insn->arg_strs,
+            (LARCH_opts.use_community_syntax && it->community_format) ? it->community_format : it->format,
+            insn->arg_strs,
             loongarch_args_parser_can_match_arg_helper, insn);
           if (insn->all_match && !(it->include && !*it->include) &&
               !(it->exclude && *it->exclude))
@@ -641,7 +676,7 @@ get_loongarch_opcode (struct loongarch_cl_insn *insn)
             }
           it++;
         }
-      while (it->name && strcasecmp (it->name, insn->name) == 0);
+      while (it->name && strcasecmp (LARCH_opts.use_community_syntax ? it->community_name : it->name, insn->name) == 0);
     }
 }
 
@@ -672,7 +707,19 @@ check_this_insn_before_appending (struct loongarch_cl_insn *ip)
             (ip->insn_bin & 0xff800000) == 0x00800000))
     {
       /* for bstr(ins|pick).[wd] */
-      if (ip->args[2] < ip->args[3])
+      int msb, lsb;
+      if (LARCH_opts.use_community_syntax)
+        {
+          lsb = ip->args[2];
+          msb = ip->args[3];
+        }
+      else
+        {
+          msb = ip->args[2];
+          lsb = ip->args[3];
+        }
+
+      if (msb < lsb)
         as_fatal (_ ("bstr(ins|pick).[wd] require msbd >= lsbd"));
     }
   else if (ip->insn->mask != 0 && (ip->insn_bin & 0xfe0003c0) == 0x04000000 &&

include/opcode/loongarch.h

@@ -37,10 +37,15 @@ extern "C"
 #define LARCH_INSN_OPC(insn) ((insn & 0xf0000000) >> 28)
     const char *const name;
 
-    /* 
+    /* The instruction name as given by the
+     * https://github.com/loongson-community/loongarch-opcodes repo.
+     * NULL if the name is identical to the manual version.  */
+    const char *const community_name;
+
+    /*
      ACTUAL PARAMETER:
 
-  // BNF with regular expression. 
+  // BNF with regular expression.
 args : token* end
 
   // just few char separate 'iden'
@@ -63,7 +68,7 @@ That 'sr' means the instruction may need relocate. '10:16' means bit field of in
 In a 'format', every 'escape's can be replaced to 'iden' or 'regname' acrroding to its meaning.
 We fill all information needed by disassembing and assembing to 'format'.
 
-  // BNF with regular expression. 
+  // BNF with regular expression.
 format : escape (literal+ escape)* literal* end
 | (literal+ escape)* literal* end
 
@@ -80,8 +85,8 @@ escape : esc_ch bit_field '<' '<' dec2
 | esc_ch bit_field
 | esc_ch    // for MACRO. non-macro format must indicate 'bit_field'
 
-  // '|' means to concatenate nonadjacent bit fields 
-  // For example, "10:16|0:4" means 
+  // '|' means to concatenate nonadjacent bit fields
+  // For example, "10:16|0:4" means
   // "16 bits starting from the 10th bit concatenating with 4 bits starting from the 0th bit".
   // This is to say "[25..10]||[3..0]" (little endian).
 b_field : dec2 ':' dec2
@@ -102,11 +107,16 @@ dec2 : [1-9][0-9]?
 */
     const char *const format;
 
+    /* The same as above, but adhering to the format and ordering as specified
+     * by the https://github.com/loongson-community/loongarch-opcodes repo.
+     * NULL if the two syntaxes have the same format. */
+    const char *const community_format;
+
     /*
 MACRO: Indicate how a macro instruction expand for assembling.
 The main is to replace the '%num'(means the 'num'th 'escape' in 'format') in 'macro' string to get the real instruction.
 
-Maybe need 
+Maybe need
 */
     const char *const macro;
     const int *include;
@@ -131,6 +141,7 @@ Maybe need
 
     /* for GAS to create hash table. */
     struct htab *name_hash_entry;
+    struct htab *community_name_hash_entry;
   };
 
   extern int is_unsigned (const char *);
@@ -206,6 +217,8 @@ Maybe need
 
     int abi_is_lp32;
     int abi_is_lp64;
+
+    int use_community_syntax;
   } LARCH_opts;
 
   extern size_t loongarch_insn_length (insn_t insn);

opcodes/loongarch-coder.c

@@ -459,6 +459,10 @@ loongarch_expand_macro_with_format_map (
   const char *src;
   char *dest;
   char buffer[8192];
+  int old_syntax_flag;
+
+  old_syntax_flag = LARCH_opts.use_community_syntax;
+  LARCH_opts.use_community_syntax = 0;
 
   if (format)
     loongarch_parse_format (format, esc1s, esc2s, bit_fields);
@@ -496,6 +500,9 @@ loongarch_expand_macro_with_format_map (
       *dest++ = *src++;
 
   *dest = '\0';
+
+  LARCH_opts.use_community_syntax = old_syntax_flag;
+
   return strdup (buffer);
 }