c12659ba51
Introducing managed pointer class ecma_pointer_t, using it to store pointer in ecma_value_t.
2015-02-03 18:25:38 +03:00
9cdc50c3ec
Disabling Wattributes warnings (because always_inline is considered incorrect without inline specifier even in LTO mode); removing used attribute from always_inline functions; removing inline specifier from ecma_lcache_lookup and ecma_stack_top_value.
2015-02-03 18:05:17 +03:00
3cb0b0a379
Revert "Introducing managed pointers classes (ecma_generic_ptr_t, ecma_pointer_t<T>); using ecma_generic_ptr_t to store pointer in ecma_value_t."
...
This reverts commit f88c0d665b4ff7e2add495bc2980adf7a1622f31.
2015-02-03 16:50:21 +03:00
b0e4d2ece1
Introducing managed pointers classes (ecma_generic_ptr_t, ecma_pointer_t<T>); using ecma_generic_ptr_t to store pointer in ecma_value_t.
2015-02-03 15:11:44 +03:00
55caaf7a88
Storing values in ecma_value_t in unpacked form (type + simple value / pointer to the heap).
2015-02-03 14:52:37 +03:00
f37a5085f6
Converting ecma_completion_value_t to class derived from ecma_value_t with additional field with for completion type and related operations.
...
Completion value are now returned through ecma_completion_value_t &ret_value argument.
2015-01-30 23:16:45 +03:00
a1f95048ae
Removing label completion values.
2015-01-30 23:00:42 +03:00
00afd4e0e2
Converting ecma_value_t to class that is used for on-stack storage of ecma-values.
...
Original ecma_value_t is renamed to ecma_value_packed_t and is used for on-heap storage.
Copy and move constructors, move assignment operator of ecma_value_t are not created.
New ecma-value return convention is introduced: ecma-values are now returned through ecma_value_t &ret_val argument.
2015-01-30 22:53:06 +03:00
4cd7f96acc
Suppressing 'member variable is not assigned in operator =' cppcheck warning, because of false positives.
2015-01-29 14:40:31 +03:00
94cb1c0e67
Changing jerry_ref_unused_variables signature to take values by const reference.
2015-01-28 19:58:20 +03:00
8b31156b5a
Forbidding use of rbp for register allocation on x86-64 platform.
...
It is necessary because of bug in gcc / g++:
Pointers allocated on rbp can be (are?) supposed to be aligned indepedently of what they actually contain.
Turning off allocation on rbp is not critical for now, because, currently, x86-64 is just development platform.
2015-01-28 17:43:04 +03:00
0f487a0bb7
Returning ecma_value_t instead of ecma_completion_value_t from ECMA_TRY_CATCH for case when completion value is normal.
2015-01-27 22:07:26 +03:00
f231b9a0d2
Moving ecma_reference_t declaration to ecma-reference.h.
2015-01-27 17:18:53 +03:00
d1925ab882
Using static_cast instead of C-style cast in MEM_DEFINE_LOCAL_ARRAY.
2015-01-27 15:40:46 +03:00
57f645c18c
Passing ecma_value_t arguments by const reference instead of by value.
2015-01-23 20:22:21 +03:00
11e37ad7f3
Switching cppcheck to c++11-compatible mode.
2015-01-22 18:07:56 +03:00
005b5370fc
Switching to g++ and corresponding changes according to C++ requirements.
2015-01-16 17:27:02 +03:00
557df54f9d
Registering temporary variables arrays in ecma-stack frames, starting mark traverse from them during GC.
2015-01-15 18:45:37 +03:00
eef966869a
Revert "Switching get_variable_value to returning value through ecma-stack (stack convention)."
...
This reverts commit 6415130925187a877b83b14db40cade0c336ea44.
2015-01-15 17:26:08 +03:00
e570d18936
Switching get_variable_value to returning value through ecma-stack (stack convention).
2015-01-14 20:09:24 +03:00
e70626f92d
Removing free of throw completion value in ECMA_TRY_CATCH macro, instead just copying it to the return value variable.
2015-01-14 18:14:28 +03:00
917ebe4a83
Creating ecma-stack frames in run_int_from_pos.
2015-01-12 20:52:04 +03:00
b843167b91
Introducing ecma-stack structures and routines.
2015-01-12 19:33:04 +03:00
d03b2dafba
Upon receiving 'give some memory back' request check if any corresponding callback set instead of asserting that.
2015-01-12 19:07:35 +03:00
f96770a811
Adding '-min' option to tools/perf.sh: the option switches to output of minimum instead of average of performance measurements.
2014-12-19 21:37:26 +03:00
42c44830ce
Removing trailing whitespace from sources.
2014-12-19 14:35:17 +03:00
6bb39bb8ea
Introducing 'try to give memory back' callback for heap allocator to use upon allocation request that can not be satisfied by the allocator.
2014-12-18 22:25:22 +03:00
8febd2bae8
Increasing heap size from 8K to 16K for MCU targets.
2014-12-18 17:17:08 +03:00
5539afc60e
Adding #pragma GCC optimize "O0" to libperipherals' modules.
2014-12-18 17:16:33 +03:00
5c4c47514a
Fixing new line marks output in tools/generator.sh.
2014-12-18 16:14:31 +03:00
84294018a3
Replacing dynamically sized on-stack arrays with heap allocated arrays defined through MEM_DEFINE_LOCAL_ARRAY.
2014-12-18 13:56:45 +03:00
1e0eea3d73
Returning NULL from mem_heap_alloc_block if requested block size is zero.
2014-12-18 13:26:42 +03:00
9b1fff1d8b
Introducing MEM_DEFINE_LOCAL_ARRAY and MEM_FINALIZE_LOCAL_ARRAY to replace on-stack allocation of dynamic sized arrays.
2014-12-17 22:18:32 +03:00
005fcb008e
Changing dynamic sized (syntactically) array to constant-sized array in ecma_op_general_object_default_value.
2014-12-17 22:14:33 +03:00
14620e30a6
Defining constant limit on magic string length instead of runtime calculation of maximum among magic strings' lengths.
2014-12-17 22:14:04 +03:00
c9b73da82d
Changing dynamic sized (syntactically) array to constant-sized array in ecma_gc_run.
2014-12-17 21:29:45 +03:00
001f262590
Limiting maximum command line arguments by CONFIG_JERRY_MAX_COMMAND_LINE_ARGS configuration option.
2014-12-17 21:29:17 +03:00
dd3271e35d
Changing uint8_t* to void* in arguments and return values of mem_heap_alloc_block, mem_heap_free_block and mem_heap_try_resize_block operations.
2014-12-17 21:27:02 +03:00
8349ad8ecf
Inter-commit performance comparison script.
2014-12-17 19:16:16 +03:00
d836dc32af
Introducing ECMA_OP_TO_NUMBER_TRY_CATCH and ECMA_OP_TO_NUMBER_FINALIZE macroses.
...
- the ECMA_OP_TO_NUMBER_TRY_CATCH macro gets number from given value,
converting the value to number if it's type is different,
and catching possible conversion exceptions;
- using the macroses instead of ecma_op_to_number to reduce allocator invocations
in several routines with arguments that are likely to be numbers.
2014-12-15 23:08:34 +03:00
386a530d4d
Upon assignment of a number to a register variable reusing the variable's space if it already contains a number.
2014-12-15 23:08:34 +03:00
eb80e817d8
Removing trailing whitespace in src/libcoreint/opcodes.c.
2014-12-15 23:08:34 +03:00
7e7bdc2aac
Rework literal indexes hash table to simple array in order to improve performance
2014-12-15 22:56:22 +03:00
864bd3de06
Fix comment.
2014-12-15 22:47:41 +03:00
ddac13f944
Fix parsing of empty file
2014-12-15 17:10:04 +03:00
c1ebb8db50
Fix expression statement parsing in case of statement starts with keyword
2014-12-12 16:36:18 +03:00
6aee69d6df
Fix native_call opcode dumping
2014-12-12 16:34:59 +03:00
4b7871bdde
Fix 'op=' assignment bytecode dumping in case of arrays
2014-12-10 19:20:33 +03:00
dc8ab27900
Split parser into parser itself, opcodes dumper and syntax errors checker.
...
Add internal hash map of literal indexes:
In this hash map key is pair of block number and literal's unique identifier in the block, and the value is a literal index that unique in the whole program.
Block is a continues array of opcodes. So, bytecode is splitted into blocks.
Each block has its own uid counter. To get literal index the interpreter looks up it in the hash map.
Thus, now JS program is able to have more than 255 identifiers/string literals.
The first 128 (0-127) uids are reserved for block's uid counter, the other 128 (128-255) are reserved for tmp variables.
2014-12-10 18:31:59 +03:00
fc9e83d290
Remove obsolete unit tests.
2014-12-10 18:29:43 +03:00
Ruben Ayrapetyan