YourWishes/jerryscript
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Parser optimizations.

Browse Source 
- parser is now non-recursive (i.e. parse function is not called recursively in any case);
 - byte-code is now more compact:
    - constants are now not immediately dumped upon occurence, but later - where necessary;
    - assignments are combined with unary / binary operations;
    - binary operations are encoded more compactly in many cases;
 - byte-code arrays are now allocated separately for each scope (so, GC of the scopes now becomes possible);
 - byte-code is dumped directly into corresponding byte-code arrays:
   - linked lists of op_meta are not now used for main code of a scope.

JerryScript-DCO-1.0-Signed-off-by: Ruben Ayrapetyan r.ayrapetyan@samsung.com
JerryScript-DCO-1.0-Signed-off-by: Andrey Shitov a.shitov@samsung.com
This commit is contained in:
Andrey Shitov
2015-11-03 19:14:19 +03:00
committed by Ruben Ayrapetyan
parent b1de93abd6
commit 50d124bfc3
51 changed files with 9044 additions and 7401 deletions
Download Patch File Download Diff File Expand all files Collapse all files
jerry-core/parser/js/collections/lit-id-hash-table.cpp
+30 -15
View File
@@ -84,11 +84,12 @@ lit_id_hash_table_free (lit_id_hash_table *table_p) /**< table's header */
} /* lit_id_hash_table_free */
/**
* Register pair in the hash table
* Register literal in the hash table
*
* @return corresponding idx
*/
void
vm_idx_t
lit_id_hash_table_insert (lit_id_hash_table *table_p, /**< table's header */
vm_idx_t uid, /**< value of byte-code instruction's argument */
vm_instr_counter_t oc, /**< instruction counter of the instruction */
lit_cpointer_t lit_cp) /**< literal identifier */
{
@@ -101,8 +102,31 @@ lit_id_hash_table_insert (lit_id_hash_table *table_p, /**< table's header */
table_p->buckets[block_id] = table_p->raw_buckets + table_p->current_bucket_pos;
}
table_p->buckets[block_id][uid] = lit_cp;
table_p->current_bucket_pos++;
lit_cpointer_t *raw_bucket_iter_p = table_p->raw_buckets + table_p->current_bucket_pos;
JERRY_ASSERT (raw_bucket_iter_p >= table_p->buckets[block_id]);
ssize_t bucket_size = (raw_bucket_iter_p - table_p->buckets[block_id]);
int32_t index;
for (index = 0; index < bucket_size; index++)
{
if (table_p->buckets[block_id][index].packed_value == lit_cp.packed_value)
{
break;
}
}
if (index == bucket_size)
{
JERRY_ASSERT ((uint8_t *) (table_p->raw_buckets + table_p->current_bucket_pos) < (uint8_t *) (table_p->buckets));
table_p->buckets[block_id][index] = lit_cp;
table_p->current_bucket_pos++;
}
JERRY_ASSERT (index <= VM_IDX_LITERAL_LAST);
return (vm_idx_t) index;
} /* lit_id_hash_table_insert */
/**
@@ -194,17 +218,8 @@ lit_id_hash_table_dump_for_snapshot (uint8_t *buffer_p, /**< buffer to dump to *
{
lit_cpointer_t lit_cp = table_p->buckets[block_index][block_idx_pair_index];
uint32_t lit_index;
for (lit_index = 0; lit_index < literals_num; lit_index++)
{
if (lit_map_p[lit_index].literal_id.packed_value == lit_cp.packed_value)
{
break;
}
}
JERRY_ASSERT (lit_index < literals_num);
uint32_t offset = bc_find_lit_offset (lit_cp, lit_map_p, literals_num);
uint32_t offset = lit_map_p[lit_index].literal_offset;
if (!jrt_write_to_buffer_by_offset (buffer_p, buffer_size, in_out_buffer_offset_p, offset))
{
return 0;