Reduce call to strlen when possible

[lygstate]

Closed: #4978
Closed: #4979

• Improve jerry_string_sz only accept UTF-8 string(that's a rare case accept CESU-8 in c/cpp code)
• The document about jerry_string_sz only support ASCII(the fact is CESU-8 before this MR), update it to support UTF-8 after this MR
• Improve all _sz function to take jerry_value_t that can construct from jerry_string_sz
• Improve JERRY_ZSTR_ARG can only accept string literal(that's UTF-8)
• Add function jerry_value_list_free to free a list of jerry_value_t
• All call to jerry_string/jerry_string_cesu8 in core indeed are removed, so when there is no linkage to it, code size is saved

The prototype of new/improved function/macros is:

jerry_value_t jerry_string_cesu8 (const jerry_char_t *buffer_p, jerry_size_t buffer_size);
jerry_value_t jerry_string_utf8 (const jerry_char_t *buffer_p, jerry_size_t buffer_size);
#define jerry_string_sz(str) jerry_string_utf8 (JERRY_ZSTR_ARG (str))

jerry_value_t jerry_error_sz (jerry_error_t error_type, const jerry_value_t message_sz);
jerry_value_t jerry_regexp_sz (const jerry_value_t pattern_sz, uint16_t flags);
jerry_value_t jerry_object_delete_sz (const jerry_value_t object, const jerry_value_t key_sz);
jerry_value_t jerry_object_get_sz (const jerry_value_t object, const jerry_value_t key_sz);
jerry_value_t jerry_object_has_sz (const jerry_value_t object, const jerry_value_t key_sz);
jerry_value_t jerry_object_set_sz (jerry_value_t object, const jerry_value_t key_sz, const jerry_value_t value);

Rename jerry_port_log to jerry_port_log_buffer

Add buffer_size parameter for function jerry_port_log_buffer, so that jerry_port_log_buffer
won't need calculate strlen when printing

JerryScript-DCO-1.0-Signed-off-by: Yonggang Luo [email protected]

[zherczeg]

What about keeping jerry_string_sz? Maybe adding jerry_string_utf8_sz()? I suspect cesu8/ascii is the same since ascii is 7 bit.

[lygstate]

What about keeping jerry_string_sz? Maybe adding jerry_string_utf8_sz()? I suspect cesu8/ascii is the same since ascii is 7 bit.

jerry_string_sz is equal to jerry_sz_cesu8 and jerry_string_cesu8 before this MR,
we have jerry_sz_utf8 and jerry_string_utf8 that have the feature of jerry_string_utf8_sz after this MR

jerry_sz_cesu8 is a macro for literal,
jerry_string_cesu8 is a function for ptr,length pair

[lygstate]

I suspect cesu8/ascii is the same since ascii is 7 bit.
I have the willing to improve string #5182

So that refactoring ASCII out is a thing because for most/(maybe all) internal string is ASCII, and for ASCII the check function is much simpler.

The macro give us future optimize string literal store, as jerry_sz_ascii declered string can be in literal store

[zherczeg]

Cesu8 is as efficient as ascii for checking, since the cpu branch predictor learns the simplest case. Two functions actually increase code size.

[lygstate]

Cesu8 is as efficient as ascii for checking, since the cpu branch predictor learns the simplest case. Two functions actually increase code size.

OK, I do the following to achieve the same effect, to code changes reduced a lot

• Improve jerry_string_sz only accept UTF-8 string(that's a rare case accept CESU-8 in c/cpp code)
• The document about jerry_string_sz only support ASCII(the fact is CESU-8 before this MR), update it to support UTF-8 after this MR

In summary jerry_string_sz preserved