kestrelcomputer / kestrel Goto Github PK

The Kestrel is a family of home-made computers, built as much as possible on open-source technology, and supporting as much as possible the open-source philosophy.

Home Page: http://kestrelcomputer.github.io/kestrel

License: Mozilla Public License 2.0

Makefile 0.21% Forth 27.67% C 4.76% Shell 0.39% Verilog 19.39% Assembly 41.42% Go 0.17% Python 2.82% M4 3.02% SystemVerilog 0.14%

kestrel freedom assembly forth verilog homebrew computer

kestrel's People

Contributors

Stargazers

Watchers

Forkers

tweakoz uho 8l miltonhill sam-falvo dfischer quinndiggity simrit1 manelpin

kestrel's Issues

Target compiler: Decimal numbers

Headers must link to previously defined word in the context.

#10

Create $NXT inner interpreter logic

Allocate and maintain a data and return stack.

Receive a line of input from the user.

Implement Machine Language Opcodes

NOP, LIT, FWM, SWM, AND, XOR, ADD, ZGO, GO, NZGO, FBM, SBM.
#22

Create ITC binding for PrintStr (TYPE) and PrintCh (EMIT)

Target compiler: Hex numbers

I'm so tired of typing #, all the damn time. :)

Embed word name in target dictionary.

Child of #10 .

This issue is to serve as an announcement to the project followers -- I'm going to be trying a new work-flow in order to familiarize myself with an online service called Huboard. I hope that my activity here won't spam people's inboxes. But, in case it does, at least you'll know why.

Target compiler: Host version of FIND

I may be able to get around needing such a definition. We'll see.

IF, THEN, EXIT, BEGIN, and AGAIN requires implementation.

child of #10 . Many of these depend on inner-interpreter primitives, which in turn requires the inline assembler.

Header order should be CFA, NFA, LFA, not LFA, NFA, CFA.

CFA must appear first to maximize runtime performance. Otherwise, NEXT will require the time and space overhead for adding an offset to the next word pointer. #10

; must terminate :-definition.

Child of #10

: must create target definition only in TARGET context.

Child of #10

Create EXECUTE

T-@ and T-! must match data width of target processor.

Related to #10 .

Tokenize input into whitespace delimited words.

Creating ITC binding for ACCEPT

Write string comparison logic

Inline Assembler

Implement VECTOR and FIXUP

VECTOR ( "name" -- ) creates a new interrupt or reset vector. It behaves exactly as if it were defined with ENTRY.

FIXUP ( a v -- ) fixes the vector destination for v to point to a.

Child of #22

Print sign-on banner to debug port when booting up.

Write $DUP primitive

Write $DO$CONST

Write a lexer that works on the host. Neither WORD nor PARSE are suitable for my needs.

I can NOT believe I have to do this.

If I have something like:

: foo     bar baz blort ;

where a tab (\t) separates foo and bar, both 32 WORD and 32 PARSE will deliver a single token, "foo\tbar". Now, this makes a whole lot of sense if your input source is a block; but it's coming from a fucking file.

Once upon a time, ANS Forth said clearly that WORD and PARSE, if given a delimiter of a space, are to also treat other whitespace-y characters the same way. Apparently, this is no longer the case; GForth 0.7.0 no longer implements this behavior.

DAMN IT! What a fucking waste of my time! 5 hours of debugging to narrow this down, and I'm sure fixing this will uncover other bugs that I could have been fixing instead.

So not happy at the moment.

Write $DO$VAR

Needed for variables, LABELs, CREATEd words, et. al.

Implement ENTRY, LABEL, CODE in legacy asm.fs

Implement ENTRY, LABEL, CODE, and END-CODE

(action)	ENTRY	LABEL	CODE
Creates Target Head	N	Y	Y
Creates Host Head	Y	Y	Y
References...	Pushes Address	Pushes Address	Calls

LABEL and ENTRY are required to place arbitrary assembly language instructions at HERE. Note that LABEL must produce a head, just like CODE, whereas ENTRY does not. LABEL and ENTRY differs from CODE in that referencing its name pushes the word's address on the stack; it does not compile a call!
#22