necto / archtoys Goto Github PK

Some sort of ISA would be helpful.

Nucleus 2.0. Coming soon...

In the layout table there should be a fourth operation type - adress operations with 2 bytes of adress. Because, provided one-byte length only 256 cells could be adressed. But task specifications claims 2-dimensional arrays 100x100 which reqire 10000 random adressed cells of memory.

In order to following a clean RISC way, i think they should differ for integer and floating point arguments:
addi - addd
subi - subd
muli - muld
divi - divd

rem is an integer operation, so it doesn't need type specification.

Develop a notation for indexing a multi-dimensional arrays.

In order to translate text files from compiler to a binary files, understood by binary translator and interpreter.

Single dimension case.
Let it be a: array [ index ]

As I've mentioned in #1 there must be an operation for putting special data, named array's head, containing a type and a length of an array.

Convert data between operations where needed

Create a convinient interface for compiler.

Enable multidimensional arrays in syntax

Do type analisys in a AST and derive types for each load and store operation

Code itself is great, but we need some sort of demonstration of its correctness.
You could find a draft of parallel machine ISA on wiki pages.

Make a wiki page with description of all build in functions, which can be called by call instruction.

Enable a basic array support.