Rework literal storage.

The new literal storage keeps ecma strings rather than having a
custom string implementation which duplicates the string management
routines. Conversions between string implementations are eliminated
which improved the performance by 4%.

JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg zherczeg.u-szeged@partner.samsung.com

Author: zherczeg

jerry-core/ecma/base/ecma-globals.h

@@ -809,15 +809,18 @@ typedef struct
  */
 typedef enum
 {
-  ECMA_STRING_CONTAINER_LIT_TABLE, /**< actual data is in literal table */
   ECMA_STRING_CONTAINER_HEAP_ASCII_STRING, /**< actual data is on the heap as an ascii string */
   ECMA_STRING_CONTAINER_HEAP_UTF8_STRING, /**< actual data is on the heap as an utf-8 string */
   ECMA_STRING_CONTAINER_UINT32_IN_DESC, /**< actual data is UInt32-represeneted Number
                                              stored locally in the string's descriptor */
   ECMA_STRING_CONTAINER_MAGIC_STRING, /**< the ecma-string is equal to one of ECMA magic strings */
   ECMA_STRING_CONTAINER_MAGIC_STRING_EX, /**< the ecma-string is equal to one of external magic strings */
 
-  ECMA_STRING_CONTAINER__MAX = ECMA_STRING_CONTAINER_MAGIC_STRING_EX /**< maximum value */
+  ECMA_STRING_LITERAL_NUMBER, /**< a literal number which is used solely by the literal storage
+                               *   so no string processing function supports this type except
+                               *   the ecma_deref_ecma_string function. */
+
+  ECMA_STRING_CONTAINER__MAX = ECMA_STRING_LITERAL_NUMBER /**< maximum value */
 } ecma_string_container_t;
 
 /**
@@ -847,6 +850,12 @@ typedef enum
 #define ECMA_STRING_GET_CONTAINER(string_desc_p) \
   ((ecma_string_container_t) ((string_desc_p)->refs_and_container & ECMA_STRING_CONTAINER_MASK))
 
+/**
+ * Checks whether the reference counter is 1.
+ */
+#define ECMA_STRING_IS_REF_EQUALS_TO_ONE(string_desc_p) \
+  (((string_desc_p)->refs_and_container >> 3) == 1)
+
 /**
  * ECMA string-value descriptor
  */
@@ -892,6 +901,9 @@ typedef struct ecma_string_t
     /** Identifier of external magic string */
     lit_magic_string_ex_id_t magic_string_ex_id;
 
+    /** Literal number */
+    ecma_value_t lit_number;
+
     /** For zeroing and comparison in some cases */
     uint32_t common_field;
   } u;

jerry-core/ecma/base/ecma-helpers-string.c

@@ -22,10 +22,7 @@
 #include "jrt.h"
 #include "jrt-libc-includes.h"
 #include "lit-char-helpers.h"
-#include "lit-literal.h"
 #include "lit-magic-strings.h"
-#include "lit-literal-storage.h"
-#include "vm.h"
 
 /** \addtogroup ecma ECMA
  * @{
@@ -72,9 +69,6 @@ typedef struct
   uint16_t length; /* Number of characters in the string */
 } ecma_string_heap_header_t;
 
-static void
-ecma_init_ecma_string_from_lit_cp (ecma_string_t *string_p,
-                                   lit_cpointer_t lit_index);
 static void
 ecma_init_ecma_string_from_magic_string_id (ecma_string_t *string_p,
                                             lit_magic_string_id_t magic_string_id);
@@ -83,39 +77,6 @@ static void
 ecma_init_ecma_string_from_magic_string_ex_id (ecma_string_t *string_p,
                                                lit_magic_string_ex_id_t magic_string_ex_id);
 
-/**
- * Initialize ecma-string descriptor with string described by index in literal table
- */
-static void
-ecma_init_ecma_string_from_lit_cp (ecma_string_t *string_p, /**< descriptor to initialize */
-                                   lit_cpointer_t lit_cp) /**< compressed pointer to literal */
-{
-  lit_literal_t lit = lit_cpointer_decompress (lit_cp);
-
-  if (LIT_RECORD_IS_MAGIC_STR (lit))
-  {
-    ecma_init_ecma_string_from_magic_string_id (string_p,
-                                                lit_magic_literal_get_magic_str_id (lit));
-
-    return;
-  }
-
-  if (LIT_RECORD_IS_MAGIC_STR_EX (lit))
-  {
-    ecma_init_ecma_string_from_magic_string_ex_id (string_p,
-                                                   lit_magic_literal_get_magic_str_ex_id (lit));
-    return;
-  }
-
-  JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-  string_p->refs_and_container = ECMA_STRING_CONTAINER_LIT_TABLE | ECMA_STRING_REF_ONE;
-  string_p->hash = lit_charset_literal_get_hash (lit);
-
-  string_p->u.common_field = 0;
-  string_p->u.lit_cp = lit_cp;
-} /* ecma_init_ecma_string_from_lit_cp */
-
 /**
  * Initialize ecma-string descriptor with specified magic string
  */
@@ -320,21 +281,6 @@ ecma_new_ecma_string_from_number (ecma_number_t num) /**< ecma-number */
   return string_desc_p;
 } /* ecma_new_ecma_string_from_number */
 
-/**
- * Allocate new ecma-string and fill it with reference to string literal
- *
- * @return pointer to ecma-string descriptor
- */
-ecma_string_t *
-ecma_new_ecma_string_from_lit_cp (lit_cpointer_t lit_cp) /**< index in the literal table */
-{
-  ecma_string_t *string_desc_p = ecma_alloc_string ();
-
-  ecma_init_ecma_string_from_lit_cp (string_desc_p, lit_cp);
-
-  return string_desc_p;
-} /* ecma_new_ecma_string_from_lit_cp */
-
 /**
  * Allocate new ecma-string and fill it with reference to ECMA magic string
  *
@@ -502,14 +448,22 @@ ecma_deref_ecma_string (ecma_string_t *string_p) /**< ecma-string */
 
       break;
     }
-    case ECMA_STRING_CONTAINER_LIT_TABLE:
     case ECMA_STRING_CONTAINER_UINT32_IN_DESC:
     case ECMA_STRING_CONTAINER_MAGIC_STRING:
     case ECMA_STRING_CONTAINER_MAGIC_STRING_EX:
     {
       /* only the string descriptor itself should be freed */
       break;
     }
+    case ECMA_STRING_LITERAL_NUMBER:
+    {
+      ecma_fast_free_value (string_p->u.lit_number);
+      break;
+    }
+    default:
+    {
+      JERRY_UNREACHABLE ();
+    }
   }
 
   ecma_dealloc_string (string_p);
@@ -532,7 +486,6 @@ ecma_string_to_number (const ecma_string_t *str_p) /**< ecma-string */
       return ((ecma_number_t) uint32_number);
     }
 
-    case ECMA_STRING_CONTAINER_LIT_TABLE:
     case ECMA_STRING_CONTAINER_HEAP_UTF8_STRING:
     case ECMA_STRING_CONTAINER_HEAP_ASCII_STRING:
     case ECMA_STRING_CONTAINER_MAGIC_STRING:
@@ -553,9 +506,12 @@ ecma_string_to_number (const ecma_string_t *str_p) /**< ecma-string */
 
       return num;
     }
-  }
 
-  JERRY_UNREACHABLE ();
+    default:
+    {
+      JERRY_UNREACHABLE ();
+    }
+  }
 } /* ecma_string_to_number */
 
 /**
@@ -628,14 +584,6 @@ ecma_string_copy_to_utf8_buffer (const ecma_string_t *string_desc_p, /**< ecma-s
       memcpy (buffer_p, data_p, size);
       break;
     }
-    case ECMA_STRING_CONTAINER_LIT_TABLE:
-    {
-      const lit_literal_t lit = lit_get_literal_by_cp (string_desc_p->u.lit_cp);
-      JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-      size = lit_charset_literal_get_size (lit);
-      memcpy (buffer_p, lit_charset_literal_get_charset (lit), size);
-      break;
-    }
     case ECMA_STRING_CONTAINER_UINT32_IN_DESC:
     {
       const uint32_t uint32_number = string_desc_p->u.uint32_number;
@@ -762,16 +710,6 @@ ecma_string_raw_chars (const ecma_string_t *string_p, /**< ecma-string */
 
   switch (ECMA_STRING_GET_CONTAINER (string_p))
   {
-    case ECMA_STRING_CONTAINER_LIT_TABLE:
-    {
-      lit_literal_t lit = lit_get_literal_by_cp (string_p->u.lit_cp);
-      JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-      length = lit_charset_literal_get_length (lit);
-      size = lit_charset_literal_get_size (lit);
-      result_p = lit_charset_literal_get_charset (lit);
-      break;
-    }
     case ECMA_STRING_CONTAINER_HEAP_UTF8_STRING:
     {
       const ecma_string_heap_header_t *data_p = ECMA_GET_NON_NULL_POINTER (ecma_string_heap_header_t,
@@ -861,11 +799,6 @@ ecma_compare_ecma_strings_longpath (const ecma_string_t *string1_p, /* ecma-stri
   {
     switch (ECMA_STRING_GET_CONTAINER (string1_p))
     {
-      case ECMA_STRING_CONTAINER_LIT_TABLE:
-      {
-        JERRY_ASSERT (string1_p->u.lit_cp != string2_p->u.lit_cp);
-        return false;
-      }
       case ECMA_STRING_CONTAINER_MAGIC_STRING:
       {
         JERRY_ASSERT (string1_p->u.magic_string_id != string2_p->u.magic_string_id);
@@ -958,13 +891,6 @@ ecma_compare_ecma_strings_longpath (const ecma_string_t *string1_p, /* ecma-stri
 
     utf8_string1_p = data_p;
   }
-  else if (ECMA_STRING_GET_CONTAINER (string1_p) == ECMA_STRING_CONTAINER_LIT_TABLE)
-  {
-    const lit_literal_t lit = lit_get_literal_by_cp (string1_p->u.lit_cp);
-    JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-    utf8_string1_p = (lit_utf8_byte_t *) lit_charset_literal_get_charset (lit);
-  }
   else
   {
     utf8_string1_p = (lit_utf8_byte_t *) jmem_heap_alloc_block ((size_t) strings_size);
@@ -988,13 +914,6 @@ ecma_compare_ecma_strings_longpath (const ecma_string_t *string1_p, /* ecma-stri
 
     utf8_string2_p = data_p;
   }
-  else if (ECMA_STRING_GET_CONTAINER (string2_p) == ECMA_STRING_CONTAINER_LIT_TABLE)
-  {
-    const lit_literal_t lit = lit_get_literal_by_cp (string2_p->u.lit_cp);
-    JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-    utf8_string2_p = (lit_utf8_byte_t *) lit_charset_literal_get_charset (lit);
-  }
   else
   {
     utf8_string2_p = (lit_utf8_byte_t *) jmem_heap_alloc_block ((size_t) strings_size);
@@ -1088,14 +1007,6 @@ ecma_compare_ecma_strings_relational (const ecma_string_t *string1_p, /**< ecma-
     utf8_string1_p = data_p;
     utf8_string1_size = (lit_utf8_size_t) string1_p->u.ascii_string.size;
   }
-  else if (ECMA_STRING_GET_CONTAINER (string1_p) == ECMA_STRING_CONTAINER_LIT_TABLE)
-  {
-    const lit_literal_t lit = lit_get_literal_by_cp (string1_p->u.lit_cp);
-    JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-    utf8_string1_p = (lit_utf8_byte_t *) lit_charset_literal_get_charset (lit);
-    utf8_string1_size = (lit_utf8_size_t) lit_charset_literal_get_size (lit);
-  }
   else
   {
     utf8_string1_size = ecma_string_get_size (string1_p);
@@ -1129,14 +1040,6 @@ ecma_compare_ecma_strings_relational (const ecma_string_t *string1_p, /**< ecma-
     utf8_string2_p = data_p;
     utf8_string2_size = (lit_utf8_size_t) string2_p->u.ascii_string.size;
   }
-  else if (ECMA_STRING_GET_CONTAINER (string2_p) == ECMA_STRING_CONTAINER_LIT_TABLE)
-  {
-    const lit_literal_t lit = lit_get_literal_by_cp (string2_p->u.lit_cp);
-    JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-    utf8_string2_p = (lit_utf8_byte_t *) lit_charset_literal_get_charset (lit);
-    utf8_string2_size = (lit_utf8_size_t) lit_charset_literal_get_size (lit);
-  }
   else
   {
     utf8_string2_size = ecma_string_get_size (string2_p);
@@ -1182,12 +1085,6 @@ ecma_string_get_length (const ecma_string_t *string_p) /**< ecma-string */
 {
   switch (ECMA_STRING_GET_CONTAINER (string_p))
   {
-    case ECMA_STRING_CONTAINER_LIT_TABLE:
-    {
-      lit_literal_t lit = lit_get_literal_by_cp (string_p->u.lit_cp);
-      JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-      return lit_charset_literal_get_length (lit);
-    }
     case ECMA_STRING_CONTAINER_MAGIC_STRING:
     {
       JERRY_ASSERT (ECMA_STRING_IS_ASCII (lit_get_magic_string_utf8 (string_p->u.magic_string_id),
@@ -1229,13 +1126,6 @@ ecma_string_get_size (const ecma_string_t *string_p) /**< ecma-string */
 {
   switch (ECMA_STRING_GET_CONTAINER (string_p))
   {
-    case ECMA_STRING_CONTAINER_LIT_TABLE:
-    {
-      lit_literal_t lit = lit_get_literal_by_cp (string_p->u.lit_cp);
-      JERRY_ASSERT (LIT_RECORD_IS_CHARSET (lit));
-
-      return lit_charset_literal_get_size (lit);
-    }
     case ECMA_STRING_CONTAINER_MAGIC_STRING:
     {
       return lit_get_magic_string_size (string_p->u.magic_string_id);

jerry-core/ecma/base/ecma-helpers.c

@@ -1429,14 +1429,14 @@ ecma_bytecode_deref (ecma_compiled_code_t *bytecode_p) /**< byte code pointer */
 
   if (bytecode_p->status_flags & CBC_CODE_FLAGS_FUNCTION)
   {
-    lit_cpointer_t *literal_start_p = NULL;
+    jmem_cpointer_t *literal_start_p = NULL;
     uint32_t literal_end;
     uint32_t const_literal_end;
 
     if (bytecode_p->status_flags & CBC_CODE_FLAGS_UINT16_ARGUMENTS)
     {
       uint8_t *byte_p = (uint8_t *) bytecode_p;
-      literal_start_p = (lit_cpointer_t *) (byte_p + sizeof (cbc_uint16_arguments_t));
+      literal_start_p = (jmem_cpointer_t *) (byte_p + sizeof (cbc_uint16_arguments_t));
 
       cbc_uint16_arguments_t *args_p = (cbc_uint16_arguments_t *) bytecode_p;
       literal_end = args_p->literal_end;
@@ -1445,7 +1445,7 @@ ecma_bytecode_deref (ecma_compiled_code_t *bytecode_p) /**< byte code pointer */
     else
     {
       uint8_t *byte_p = (uint8_t *) bytecode_p;
-      literal_start_p = (lit_cpointer_t *) (byte_p + sizeof (cbc_uint8_arguments_t));
+      literal_start_p = (jmem_cpointer_t *) (byte_p + sizeof (cbc_uint8_arguments_t));
 
       cbc_uint8_arguments_t *args_p = (cbc_uint8_arguments_t *) bytecode_p;
       literal_end = args_p->literal_end;

jerry-core/ecma/base/ecma-helpers.h

@@ -19,7 +19,6 @@
 
 #include "ecma-globals.h"
 #include "jmem-allocator.h"
-#include "lit-cpointer.h"
 #include "lit-strings.h"
 
 /** \addtogroup ecma ECMA
@@ -165,7 +164,6 @@ extern ecma_string_t *ecma_new_ecma_string_from_utf8 (const lit_utf8_byte_t *, l
 extern ecma_string_t *ecma_new_ecma_string_from_code_unit (ecma_char_t);
 extern ecma_string_t *ecma_new_ecma_string_from_uint32 (uint32_t);
 extern ecma_string_t *ecma_new_ecma_string_from_number (ecma_number_t);
-extern ecma_string_t *ecma_new_ecma_string_from_lit_cp (lit_cpointer_t);
 extern ecma_string_t *ecma_new_ecma_string_from_magic_string_id (lit_magic_string_id_t);
 extern ecma_string_t *ecma_new_ecma_string_from_magic_string_ex_id (lit_magic_string_ex_id_t);
 extern ecma_string_t *ecma_concat_ecma_strings (ecma_string_t *, ecma_string_t *);

jerry-core/ecma/base/ecma-init-finalize.c

@@ -19,6 +19,7 @@
 #include "ecma-init-finalize.h"
 #include "ecma-lcache.h"
 #include "ecma-lex-env.h"
+#include "ecma-literal-storage.h"
 #include "jmem-allocator.h"
 
 /** \addtogroup ecma ECMA
@@ -37,6 +38,7 @@ ecma_init (void)
   ecma_gc_init ();
   ecma_init_builtins ();
   ecma_lcache_init ();
+  ecma_init_lit_storage ();
   ecma_init_environment ();
 
   jmem_register_free_unused_memory_callback (ecma_free_unused_memory);
@@ -53,6 +55,7 @@ ecma_finalize (void)
   ecma_finalize_environment ();
   ecma_finalize_builtins ();
   ecma_gc_run ();
+  ecma_finalize_lit_storage ();
 } /* ecma_finalize */
 
 /**