Currently we keep all local variables in the scope chain as long as
a nested function is alive. These variables consume memory although
several of them are never used. This patch focuses on removing those
unused ones whose can be detected at parsing time.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Now the debugger generates a breakpoint for each function before
its first executable statement. This allows inspecting the function
arguments. Position (line and column) info is also added which
simplifies finding of anonymus functions. Several tests were
update to check more corner cases.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
For example Number(".") and Number("e5") should be NaN not
zero. Parsing Number("e") caused buffer overflow as well.
Infinity parsing is a bit faster now.
Fixes#1636.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Some internal properties are incorrectly handled as objects and
marked as visited. This memory overwrite caused random crashes
in IoT.js.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
This makes the code more robust and error prone, since certain
messages cannot be received in certain situations, e.g. an eval
command during garbage collection.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Besides the breakpoint list the JavaScript source code is
also sent to the debugger. This feature allows debugging
eval() function calls.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Two issues were fixed as well: inserting breakpoints before non-directive
prologue strings and the receive can process multiple messages.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The server can accept a string, execute it with eval,
and returns with the result converted to string. In
case of exception it returns with the exception message.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
In a previous patch the original null count behaviour of the
property hashmap is just only partially restored. Now it is
restored fully Also another assertion was forgot to be fixed
in another patch, and we do so now.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Because of the fallthrough case the value of 3 is also allowed in the default case.
Fixes#1555.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Triggering hashmap recreate during property delete may damage the
property chain list, so recreate postponed after the delete is done.
Next attempt to fix#1533.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
There are two types of NULL values in the property hashmap: deleted
entries, and never used entries. The current implementation tracks
only the never used entries, and never scales back the hashmap.
After this patch property delete can also recreate the hashmap, or
remove it entirely if there are too few items in the array.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The regression-test-issue-736.js is also changed, since it expects
throwing an error, which does not happen anymore because the "x"
variable is undefined.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Rework date to string conversion, and remove nan checks from inline
functions. Furtermore some inline functions are changed to normal
functions.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Although many Math built-in functions have a similar structure
these code paths were implemented as separate C functions.
After this patch only one common dispatcher remains which shares
the common code paths of different built-in functions. This
reduces the binary size by 1 Kbyte.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Before this patch each built-in routine had a separate C
function. Although these functions shared a lot of common
operations compilers cannot eliminate these code duplications
in the final binary code. This patch introduces a single C
function which handles all date prototype routines. Each
shared code block has only one binary represenation.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The JSON parser required a zero terminated writable copy of
the original string. The requirement is eliminated from the
project to reduce peak memory consumption.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Property names were always required a string reference which consumed
a large amount of memory for arrays. This patch reduces this consumption
by directly storing the value part of certain strings.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
This change allows easier access to array length which improves
the performance of inserting new items into an array and there
is no need to allocate length strings anymore. The trade-of is
that array length cannot be cached anymore.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Free a new bit in the property descriptor by combining internal
and special property types into one group. Also simplify checking
special properties since bit-and operation is not needed anymore.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
This patch reduces memory consumption for strings such as "0" or "123"
by 8 bytes and "4294967295" by 16 bytes. The hash computation is changed
for using the lower 16 bits for these strings which is much faster than
converting the value to string first and compute the hash. The trade-of
is a small overhead when strings are created or concatenated.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Properties are changed to a type and value pair instead of a pointer to an internal
representation. Functions such as ecma_op_object_get_[own_]property do not
return with property pointers anymore.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The ecma_op_object_get_[own_]property calls should be phased out from
the project eventually and virtual properties should be introduced instead.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Removing a lot of ECMA_PROPERTY_VALUE_PTR macro calls. The only drawback
is free callbacks for native objects cannot be deleted anymore. Redefining
a free callback is a rare case, so this trade-of is acceptable.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Zero out all globals (and remove unnecessary init() functions).
Move snapshot globals to a temporary stack variable.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Currently the static variables of the allocator are moved into a
global structure.
The future plan is supporting multiple context models with different
advantages and disadvantages. Users can select the most appropriate
context model for their own use case. This context model must be
selected at compile time using compiler defines.
Currently only one model is implemented, which will be the default
context model: the context is stored in global variables, so only
a single context is available which always exists.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Since upper/lower case conversions are unspecified by the standard,
we convert ASCII characters only, and a few other characters for
testing purposes. Because these are just random cases, it is better
to remove them before the release. At some point we could add a
unicode compatible case conversion which can be enabled at compile
time.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The "length" property name is the most frequently used built-in string
and also frequently created by various hot-paths. New functions are
added to improve the speed of the "length" string creation.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Allocate a single memory block for strings, rather than a separate string header
and string characters block. In the past strings were split into 8 byte chunks,
and large amount of legacy code is designed for that representation. However the
current allocator allows block allocation so we don't need those complicated
algorithms anymore. This patch is a cleanup rather than an optimization.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Short summary of the changes:
- The ecma_reference_t is removed, and its helper functions are deleted.
- The delete operation does not depend on ecma_reference_t anymore.
- A new resolve function is added which returns the current value
of a named binding in the context chain.
- The vm_op_set_value does not perform type conversions when its arguments
has the appropriate types.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
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 [email protected]
Several internal properties are removed and directly stored as
part of the object. Faster built-in and JS function processing.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Skip put result phase for opcodes which put their
result onto the stack or they don't have result at all.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The number of delete opcodes is reduced to two from six. The
range of numbers which can be included in the byte code is
doubled from (-127,127) to (-256,256).
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
or no arguments at all. To improve the performance, the branch argument
flag is changed into a get argument type which eliminates a flag check
for all executed opcodes. Furthermore branch opcodes skip the put result
phase by changing the "break" to "continue" statements.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The implementation is simplified and optimized
resulting the removal a large amount of code.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The cache stores only real properties now, because storing NULLs has
little benefit according to tests. Since only real properties are
stored now, there is no need to create real references to objects
and property names, which reduces the keeping of dead objects after
garbage collection.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
The ecma_op_to_boolean return value is changed to bool for faster
evaluation, and no need to swap operandos of relational compare.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
the three boolean arguments of ecma_create_named_data_property and the
two boolean arguments of ecma_create_named_accessor_property are combined
into one uint8_t argument. On ARM-32 it is preferred to have less than
four arguments, since these arguments can be passed in registers.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
function is removed and incorporated into ecma_op_general_object_put. Also most
properties are directly created / updated instead of using ecma_builtin_helper_def_prop.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
- Branch argument information is encoded in the vm byte
code data, so CBC flags are not loaded anymore
- The free_flags variable is removed from the vm_loop
- Two cases are removed from "get arguments", argument
processing is simplified
- The two opcode tables are merged to one
- The VM_OC_POP opcode has no result, so the break is changed to continue.
- The VM_OC_PUSH_NUMBER can use ecma_make_integer_value.
JerryScript-DCO-1.0-Signed-off-by: Zoltan Herczeg [email protected]
Zoltan Herczeg