Build your own Forth for microchip PIC (Episode 839) : Forth

I've asked about a primitive kernel to implement Forth upon. Instead of
looking at existing examples maybe a better approach is to post a
description of my current 16F NPCI bytecode and discuss. I'll try to add
typical forth stack comments for each primitive. This virtual machine
implemented a frame pointer for locals and parameters. Comments as to
the necessity of such a contruct welcome.

--------------------------------------------------------
assign ; 0 - assign ( n1 -- )

/ Badly implemented for forth address was encoded into the bytecode
/ need to reimplement where the address is on the stack. Should it
/ be ( n1 n2 -- ) with n1 being written to n2? or vice-versa?
--------------------------------------------------------
plus ; 1 - plus ( n1 n2 -- n1 + n2 )
--------------------------------------------------------
equal ; 2 - equal ( n1 n2 -- n1 == n2 )

/ All logical ops current use C style 0 for false, everything else
/ is true. In addition returns 1 on stack for true. Is there a
/ better implementation?
--------------------------------------------------------
notequal ; 3 - notequal ( n1 n2 -- n1 != n2 )
--------------------------------------------------------
minus ; 4 - minus ( n1 n2 -- n2 - n1 )
--------------------------------------------------------
bitand ; 5 - bitand ( n1 n2 -- n1 & n2 )
--------------------------------------------------------
bitor ; 6 - bitor ( n1 n2 -- n1 | n2 )
--------------------------------------------------------
bitxor ; 7 - bitxor ( n1 n2 -- n1 ^ n2 )
--------------------------------------------------------
logior ; 8 - logical or ( n1 n2 -- n1 || n2 )
--------------------------------------------------------
logiand ; 9 - logical and ( n1 n2 -- n1 && n2 )
--------------------------------------------------------
procret ; 10 - return from procedure ( n1 -- )

/ NPCI didn't have a separate return stack. So the return address
/ is pulled off the only stack available. Needs to be reimplemented
/ but fundamentally this is EXIT. Pulled return address off stack
/ and put in IP
--------------------------------------------------------
callproc ; 11 - call procedure ( -- n1 )

/ Same here. Pushed return address on stack and jumped to address
/ encoded after bytecode. Needs to be reimplemented for typical
/ FORTH inner loop. Fundamentally is ENTER.
--------------------------------------------------------
shiftleft ; 12 - shift left ( n1 n2 -- n2 << n1 )

/ No rotate. shift only
--------------------------------------------------------
shiftright ; 13 - shift right ( n1 n2 -- n2 >> n1 )

/ No rotate. shift only
--------------------------------------------------------
greater ; 14 - greater ( n1 n2 -- n2 > n1 )

/ Same as equals in terms of results
--------------------------------------------------------
less ; 15 - less than ( n1 n2 -- n2 < n1 )

/ Same as equals in terms of results
--------------------------------------------------------
greatereq ; 16 - greater than or equals ( n1 n2 -- n2 >= n1 )

/ Same as equals in terms of results
--------------------------------------------------------
lesseq ; 17 - less than or equals ( n1 n2 -- n2 <= n1 )

/ Same as equals in terms of results
--------------------------------------------------------
bitval ; 18 - Get bit value ( n1 -- n1 )

/ Needs to be reimplemented. Gets the value of bit on TOS.
/ variable address currently implemented in bytecode and needs to be
/ moved to the stack.
--------------------------------------------------------
bitassign ; 19 - Assign bit ( n1 n2 -- )

/ Needs to be reimplemented. sets the value of bit on TOS to value
// underneath in n2.
/ variable address currently implemented in bytecode and needs to be
/ moved to the stack.
--------------------------------------------------------
not ; 20 - Logical not - ( n1 -- !n1 )
// Use xor 255 for bitwise
--------------------------------------------------------
dotable ; 21 - Access table. ( n1 -- n1 )

/ probable needs reimplementation. Fetches byte from table in program
/ memory. Table base encoded in bytecode. incoming n1 is index. Outgoing
/ is value. Used for static tables. Probably can be dropped for forth.
--------------------------------------------------------
pushfpval ; 22 - Push a frame pointer val on stack ( n1 -- n1 )

/ Uses frame pointer. Not sure forth needs it
--------------------------------------------------------
pushfpref ; 23 - Push a frame pointer ref on stack ( n1 -- n1 )

/ Uses frame pointer. Not sure forth needs it
--------------------------------------------------------
assignfp ; 24 - Assign to a frame pointer variable ( n1 n2 -- )
/ Uses frame pointer. Not sure forth needs it
/ May implement bit assign better than original because address is on
/ the stack.
--------------------------------------------------------
bassignfp ; 25 - Assign to a frame pointer varaiable bit ( n1 n2 n3 -- )
/ Uses frame pointer. Not sure forth needs it
/ May implement bit assign better than original because address is on
/ the stack.
--------------------------------------------------------
link ; 26 - Save old fp on TOS and point FP to TOS ( -- n1 )
--------------------------------------------------------
unlink ; 27 - Restore FP to previous value. ( n1 -- )
--------------------------------------------------------
movetos ; 28 - move the stack pointer number of bytes
; specified by value on TOS ( n1 -- variable drop )
--------------------------------------------------------
charmode ; 29 - Set stack to 8 bit mode ( -- )
/ impements 8 bit stack unneccary for forth
--------------------------------------------------------
intmode ; 30 - Set stack to 16 bit mode
/ impements 16 bit stack unneccary for forth
--------------------------------------------------------
stackadj ; 31 - Adjust stack making second op into int.
; presumes that top op is already an int.
/ intmode/charmode conversion. unnecessary for forth
--------------------------------------------------------
numbers ; push number on the stack ( -- n1 )
--------------------------------------------------------
variables ; push contents of memory on stack ( -- n1 )

needs to be reimplemented as target address not on stack
--------------------------------------------------------
conditional goto; jump if TOS is 0 ( n1 -- )

Needs to be reimplemented because target address not on stack.
--------------------------------------------------------
goto ; jump to target ( -- )

Needs to be reimplemented because target address not on stack
--------------------------------------------------------

If anyone gets a chance, take a look and tell me what you think.
All of these are coded and tested in PIC 16F assembly.

Thanks,

BAJ

none Byron Jeff wrote:

[snipped]

Byron, could I ask you reply to your own root post? Then each of these
wouldn't end up as a seperate thread; they really belong together.

Thanks

--
Regards
Alex McDonald

In article <A7GdnZLZNr_ZMODbnZ2dnUVZ8qPinZ2d@bt.com>,
Alex McDonald <blog@rivadpm.com> wrote:
>none Byron Jeff wrote:
>
>[snipped]
>
>Byron, could I ask you reply to your own root post? Then each of these
>wouldn't end up as a seperate thread; they really belong together.

I thought of them as separate issues personally. But if that's what you
would prefer. Will do.

BAJ

none Byron Jeff wrote:
> In article <A7GdnZLZNr_ZMODbnZ2dnUVZ8qPinZ2d@bt.com>,
> Alex McDonald <blog@rivadpm.com> wrote:
>> none Byron Jeff wrote:
>>
>> [snipped]
>>
>> Byron, could I ask you reply to your own root post? Then each of these
>> wouldn't end up as a seperate thread; they really belong together.
>
> I thought of them as separate issues personally. But if that's what you
> would prefer. Will do.
>
> BAJ

A threaded newsreader makes it easier to see the flow, and will save you
(and your readers) the effort of repetition. I'm interested in what
you're doing, hence the request.

--
Regards
Alex McDonald

In article <D4-dndFfGqbKLuDbRVnyvQA@bt.com>,
Alex McDonald <blog@rivadpm.com> wrote:
>none Byron Jeff wrote:
>> In article <A7GdnZLZNr_ZMODbnZ2dnUVZ8qPinZ2d@bt.com>,
>> Alex McDonald <blog@rivadpm.com> wrote:
>>> none Byron Jeff wrote:
>>>
>>> [snipped]
>>>
>>> Byron, could I ask you reply to your own root post? Then each of these
>>> wouldn't end up as a seperate thread; they really belong together.
>>
>> I thought of them as separate issues personally. But if that's what you
>> would prefer. Will do.
>>
>> BAJ
>
>A threaded newsreader makes it easier to see the flow, and will save you
>(and your readers) the effort of repetition. I'm interested in what
>you're doing, hence the request.

I have a threaded newsreader. Again I just felt they were separate
topics of discussion and didn't want to get multiple threads muddled
together into a single tangled ball.

BAJ

none Byron Jeff wrote:
> In article <D4-dndFfGqbKLuDbRVnyvQA@bt.com>,
> Alex McDonald <blog@rivadpm.com> wrote:
>> none Byron Jeff wrote:
>>> In article <A7GdnZLZNr_ZMODbnZ2dnUVZ8qPinZ2d@bt.com>,
>>> Alex McDonald <blog@rivadpm.com> wrote:
>>>> none Byron Jeff wrote:
>>>>
>>>> [snipped]
>>>>
>>>> Byron, could I ask you reply to your own root post? Then each of these
>>>> wouldn't end up as a seperate thread; they really belong together.
>>> I thought of them as separate issues personally. But if that's what you
>>> would prefer. Will do.
>>>
>>> BAJ
>> A threaded newsreader makes it easier to see the flow, and will save you
>> (and your readers) the effort of repetition. I'm interested in what
>> you're doing, hence the request.
>
> I have a threaded newsreader. Again I just felt they were separate
> topics of discussion and didn't want to get multiple threads muddled
> together into a single tangled ball.

If they were different topics, they'd have different, more relevant
subject lines, rather than the artificial distinction of a phony number.

Cheers,
Elizabeth

--
==================================================
Elizabeth D. Rather (US & Canada) 800-55-FORTH
FORTH Inc. +1 310-491-3356
5155 W. Rosecrans Ave. #1018 Fax: +1 310-978-9454
Hawthorne, CA 90250
http://www.forth.com

"Forth-based products and Services for real-time
applications since 1973."
==================================================

On Sat, 23 Jun 2007 18:25:20 -0500, byron@upstairs.(none) (Byron Jeff)
wrote:

>I've asked about a primitive kernel to implement Forth upon. Instead of
>looking at existing examples maybe a better approach is to post a
>description of my current 16F NPCI bytecode and discuss. I'll try to add
>typical forth stack comments for each primitive. This virtual machine
>implemented a frame pointer for locals and parameters. Comments as to
>the necessity of such a contruct welcome.

If you haven't already done so, have a look at the Open Terminal
Architecture (OTA) specifications. This wa the result of a work
by Forth Inc and MPE for Europay, now part of MasterCard. OTA
is basically a two-stack VM based on Forth, extended for C, and
focussed on code density. Google tells me that you can find the
documents at:
www.forth.org.ru/~mlg/std/OTA/

Stephen


--
Stephen Pelc, stephenXXX@mpeforth.com
MicroProcessor Engineering Ltd - More Real, Less Time
133 Hill Lane, Southampton SO15 5AF, England
tel: +44 (0)23 8063 1441, fax: +44 (0)23 8033 9691
web: http://www.mpeforth.com - free VFX Forth downloads

Op Sat, 23 Jun 2007 18:25:20 -0500 schreef none:

> I've asked about a primitive kernel to implement Forth upon. Instead of
> looking at existing examples maybe a better approach is to post a
> description of my current 16F NPCI bytecode and discuss. I'll try to add
> typical forth stack comments for each primitive. This virtual machine
> implemented a frame pointer for locals and parameters. Comments as to
> the necessity of such a contruct welcome.
>
> --------------------------------------------------------
> assign ; 0 - assign ( n1 -- )
>
> / Badly implemented for forth address was encoded into the bytecode
> / need to reimplement where the address is on the stack. Should it
> / be ( n1 n2 -- ) with n1 being written to n2? or vice-versa?
> --------------------------------------------------------
> plus ; 1 - plus ( n1 n2 -- n1 + n2 )
> --------------------------------------------------------
> equal ; 2 - equal ( n1 n2 -- n1 == n2 )
>
> / All logical ops current use C style 0 for false, everything else
> / is true. In addition returns 1 on stack for true. Is there a
> / better implementation?
> --------------------------------------------------------
> notequal ; 3 - notequal ( n1 n2 -- n1 != n2 )
> --------------------------------------------------------
> minus ; 4 - minus ( n1 n2 -- n2 - n1 )

I think you have the stack reversed from the normal Forth notation
( n1 n2 -- n1-n2 )
Push n1 first, than n2, and subtracht n2 from n1.

> --------------------------------------------------------
> bitand ; 5 - bitand ( n1 n2 -- n1 & n2 )
> --------------------------------------------------------
> bitor ; 6 - bitor ( n1 n2 -- n1 | n2 )
> --------------------------------------------------------
> bitxor ; 7 - bitxor ( n1 n2 -- n1 ^ n2 )
> --------------------------------------------------------
> logior ; 8 - logical or ( n1 n2 -- n1 || n2 )
> --------------------------------------------------------
> logiand ; 9 - logical and ( n1 n2 -- n1 && n2 )
> --------------------------------------------------------
> procret ; 10 - return from procedure ( n1 -- )
>
> / NPCI didn't have a separate return stack. So the return address
> / is pulled off the only stack available. Needs to be reimplemented
> / but fundamentally this is EXIT. Pulled return address off stack
> / and put in IP
> --------------------------------------------------------
> callproc ; 11 - call procedure ( -- n1 )
>
> / Same here. Pushed return address on stack and jumped to address
> / encoded after bytecode. Needs to be reimplemented for typical
> / FORTH inner loop. Fundamentally is ENTER.
> --------------------------------------------------------
> shiftleft ; 12 - shift left ( n1 n2 -- n2 << n1 )

Reverse the stack: ( n1 n2 -- n1<<n2 )
> / No rotate. shift only
> --------------------------------------------------------
> shiftright ; 13 - shift right ( n1 n2 -- n2 >> n1 )

Idem dito. n1<<n2
> / No rotate. shift only
> --------------------------------------------------------
> greater ; 14 - greater ( n1 n2 -- n2 > n1 )
>

n1<n2

I could go on, but I'm going to eat my vanilla yoghurt ;-)

--
Coos

CHForth, 16 bit DOS applications
http://home.hccnet.nl/j.j.haak/forth.html

In article <1sxqxtculdr4m$.1ly164xd3gegw$.dlg@40tude.net>,
Coos Haak <chforth@hccnet.nl> wrote:
>Op Sat, 23 Jun 2007 18:25:20 -0500 schreef none:
>
>> I've asked about a primitive kernel to implement Forth upon. Instead of
>> looking at existing examples maybe a better approach is to post a
>> description of my current 16F NPCI bytecode and discuss.
>> minus ; 4 - minus ( n1 n2 -- n2 - n1 )
>
>I think you have the stack reversed from the normal Forth notation
>( n1 n2 -- n1-n2 )
>Push n1 first, than n2, and subtracht n2 from n1.

You are correct. My NPCI compiler converts infix to postfix
using a operator stack. Operands pass through unmodified.

I misunderstood where the TOS is. I just finished reading the gforth
manual. TOS is furthest to the right, not the left. Got it.

BTW the primitives do implement the proper stack effect. I get a D- for
listing them incorrectly. Sorry.

BAJ