prog8/docs/source/todo.rst

68 lines
2.1 KiB
ReStructuredText
Raw Normal View History

2018-08-06 01:35:43 +00:00
====
TODO
====
2019-01-24 01:43:25 +00:00
Memory Block Operations integrated in language?
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2018-08-06 01:35:43 +00:00
2019-04-11 23:06:46 +00:00
list,string memory block operations?
2018-08-06 01:35:43 +00:00
- list operations (whole list, individual element)
operations: set, get, copy (from another list with the same length), shift-N(left,right), rotate-N(left,right)
clear (set whole list to the given value, default 0)
2018-10-30 19:29:03 +00:00
- list operations ofcourse work identical on vars and on memory mapped vars of these types.
2018-08-06 01:35:43 +00:00
- strings: identical operations as on lists.
2019-04-11 23:06:46 +00:00
these should call optimized pieces of assembly code, so they run as fast as possible
2018-08-06 01:35:43 +00:00
For now, we have the ``memcopy``, ``memset`` and ``strlen`` builtin functions.
2019-01-24 01:43:25 +00:00
2019-08-11 10:29:18 +00:00
Fixes
^^^^^
fix asmsub parameters so this works::
asmsub aa(byte arg @ Y) -> clobbers() -> () {
byte local = arg ; @todo fix 'undefined symbol arg' that occurs here
A=44
}
2019-01-24 01:43:25 +00:00
More optimizations
^^^^^^^^^^^^^^^^^^
Add more compiler optimizations to the existing ones.
- on the language AST level
- on the final assembly source level
- can the parameter passing to subroutines be optimized to avoid copying?
2019-01-24 01:43:25 +00:00
2019-01-31 22:46:58 +00:00
- subroutines with 1 or 2 byte args (or 1 word arg) should be converted to asm calling convention with the args in A/Y register
this requires rethinking the way parameters are represented, simply injecting vardecls to
declare local variables for them is not always correct anymore
2019-01-31 22:46:58 +00:00
2019-08-11 08:44:58 +00:00
- working subroutine inlining (taking care of vars and identifier refs to them)
2019-01-24 01:43:25 +00:00
Also some library routines and code patterns could perhaps be optimized further
Eval stack redesign? (lot of work)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The eval stack is now a split lsb/msb stack using X as the stackpointer.
Is it easier/faster to just use a single page unsplit stack?
It could then even be moved into the zeropage to greatly reduce code size and slowness.
Or just move the LSB portion into a slab of the zeropage.
2019-01-29 22:48:26 +00:00
Allocate a fixed word in ZP that is the TOS so we can operate on TOS directly
without having to to index into the stack?
Misc
^^^^
2019-04-11 23:06:46 +00:00
- are there any other missing instructions in the code generator?