An implementation of Forth with inner interpreter using extended indirect thread code and a dictionary made of machine independent vocabularies.

Only the inner interpreter and vocabularies related to systems, drives and primitives are machine dependent.

The extended indirect thread code (EITC) gives to Forth inner interpreter the decision of when use "push and pull" or "jump and link".

More information in https://github.com/agsb/immu

I'm requesting for comments about immu, an extended indirect thread code for forth.

I welcome your patience.

On Wednesday, 10 November 2021 at 13:44:22 UTC, barcello...@gmail.com wrote:
> An implementation of Forth with inner interpreter using extended indirect thread code and a dictionary made of machine independent vocabularies.
>
> Only the inner interpreter and vocabularies related to systems, drives and primitives are machine dependent.
>
> The extended indirect thread code (EITC) gives to Forth inner interpreter the decision of when use "push and pull" or "jump and link".
>
> More information in https://github.com/agsb/immu
>
> I'm requesting for comments about immu, an extended indirect thread code for forth.
>
> I welcome your patience.

Please check and update the links here. Thank you.
https://github.com/agsb/immu/blob/main/Details.md
Quite a few I could not open from the UK

bibliography
“http://worrydream.com/refs/Moore - Forth - The Early Years.pdf”

“http://www.forth.org/POL.pdf”

“http://thinking-forth.sourceforge.net/”

“http://www.forth.org/KittPeakForthPrimer.pdf”

“http://www.forth.org/fig-forth/fig-forth_PDP-11.pdf”

"http://forth.org/OffeteStore/1013_eForthAndZen.pdf"

"https://www.bradrodriguez.com/papers/moving1.htm"

“https://muforth.nimblemachines.com/threaded-code/”

“https://muforth.nimblemachines.com/call-versus-branch-and-link/”

“https://riscv.org/specifications/”

I also had this problem, then noticed that ThunderBird does
not strip the trailing quote mark, although does the first.

Jan Coombs
--

On Wed, 10 Nov 2021 06:04:01 -0800 (PST)
Jurgen Pitaske <jpitaske@gmail.com> wrote:
>
> Please check and update the links here. Thank you.
> https://github.com/agsb/immu/blob/main/Details.md
> Quite a few I could not open from the UK
>
> bibliography
> “http://worrydream.com/refs/Moore - Forth - The Early Years.pdf”
>
> “http://www.forth.org/POL.pdf”
>
> “http://thinking-forth.sourceforge.net/”
>
> “http://www.forth.org/KittPeakForthPrimer.pdf”
>
> “http://www.forth.org/fig-forth/fig-forth_PDP-11.pdf”
>
> "http://forth.org/OffeteStore/1013_eForthAndZen.pdf"
>
> "https://www.bradrodriguez.com/papers/moving1.htm"
>
> “https://muforth.nimblemachines.com/threaded-code/”
>
> “https://muforth.nimblemachines.com/call-versus-branch-and-link/”
>
> “https://riscv.org/specifications/”

Alvaro Barcellos <barcellos.alvaro@gmail.com> writes:
>The extended indirect thread code (EITC) gives to Forth inner interpreter the decision of when use "push and pull" or "jump and link".

Can you elaborate on that? I followed the link, but a short look
around did not enlighten me on what you mean by that.

- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html
comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: http://www.forth200x.org/forth200x.html
EuroForth 2021: https://euro.theforth.net/2021

Links updated.

Some errors by quotation marks ;

Thank's for point it.

--

AVISO DE CONFIDENCIALIDADE
"Esta mensagem da CPRM é enviada exclusivamente
a seu destinatário e pode conter informações confidenciais, protegidas por
sigilo
profissional. Se você a recebeu indevidamente, a utilização desta
mensagem é desautorizada e queira, por gentileza, devolvê-la ao emitente,
esclarecendo o equívoco."

Links updated.

Some errors by "quotation mark" ;

Thank's for point it.

> Can you elaborate on that? I followed the link, but a short look
> around did not enlighten me on what you mean by that.
>
> - anton

I will try.

The proposed extended interactive thread code (EITC) is based on changing the internal interpreter and dictionary structure.

All Forth words are formed by "name" and "code", like a tuple.

Ideally, primitive words do not branch to other words and compound words just contain references to other words.

The EITC changes the internal interpreter, by wrapping the DOCOLON code also known as NEST, SEMIS also known as UNNEST and NEXT into a single routine, by modify the dictionary structure, leaving only references within the compound words, and also alters the primitive words, where DOCODE is changed to NULL (0x0).

These changes make it possible to use the NULL comparison of the first reference of each word to decide whether the inner interpreter should "jump and link" to a primitive word reference to execute code or "push to and pull from" from the return stack, the reference to a compound word.

By eliminating the need to "jump to all references", the built-in interpreter works more evident and efficiently.

This change also makes it possible for vocabularies to not have specific processor codes, so they are portable and relocatable – and can be shared directly. Only system, driver and primitive vocabulary contain specific codes.

Alvaro Barcellos <barcellos.alvaro@gmail.com> writes:
>The EITC changes the internal interpreter, by wrapping the DOCOLON code als=
>o known as NEST, SEMIS also known as UNNEST and NEXT into a single routine,=
> by modify the dictionary structure, leaving only references within the com=
>pound words, and also alters the primitive words, where DOCODE is changed t=
>o NULL (0x0).
>
>These changes make it possible to use the NULL comparison of the first refe=
>rence of each word to decide whether the inner interpreter should "jump and=
> link" to a primitive word reference to execute code or "push to and pull f=
>rom" from the return stack, the reference to a compound word.
>
>By eliminating the need to "jump to all references", the built-in interpret=
>er works more evident and efficiently.

Whether it's more efficient would have to be determined by comparing
the performance (or memory usage) against a classic ITC system. The
data I have shows that dynamically, 12-16% of NEXTs execute DOCOL, and
the same number execute SEMIS, leaving 68%-76% of NEXTs for other word
types. Special-casing DOCOL and SEMIS in NEXT incurs overhead for
these other word types that has to be amortized by more efficient
handling when a DOCOL or SEMIS occurs.

- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html
comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: http://www.forth200x.org/forth200x.html
EuroForth 2021: https://euro.theforth.net/2021

> Whether it's more efficient would have to be determined by comparing
> the performance (or memory usage) against a classic ITC system. The
> data I have shows that dynamically, 12-16% of NEXTs execute DOCOL, and
> the same number execute SEMIS, leaving 68%-76% of NEXTs for other word
> types. Special-casing DOCOL and SEMIS in NEXT incurs overhead for
> these other word types that has to be amortized by more efficient
> handling when a DOCOL or SEMIS occurs.
> - anton

Thanks by you remarks.
Could you please indicate which Forth you used for these usage values?
I'm curious to know what other types of words they could be and how they are implemented.

alvaro
> --
> M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html
> comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
> New standard: http://www.forth200x.org/forth200x.html
> EuroForth 2021: https://euro.theforth.net/2021

--

AVISO DE CONFIDENCIALIDADE
"Esta mensagem da CPRM é enviada exclusivamente
a seu destinatário e pode conter informações confidenciais, protegidas por
sigilo
profissional. Se você a recebeu indevidamente, a utilização desta
mensagem é desautorizada e queira, por gentileza, devolvê-la ao emitente,
esclarecendo o equívoco."

Em quinta-feira, 11 de novembro de 2021 às 13:52:56 UTC-3, Anton Ertl escreveu:

> >By eliminating the need to "jump to all references", the built-in interpret> >er works more evident and efficiently.
> Whether it's more efficient would have to be determined by comparing
> the performance (or memory usage) against a classic ITC system. The
> data I have shows that dynamically, 12-16% of NEXTs execute DOCOL, and
> the same number execute SEMIS, leaving 68%-76% of NEXTs for other word
> types. Special-casing DOCOL and SEMIS in NEXT incurs overhead for
> these other word types that has to be amortized by more efficient
> handling when a DOCOL or SEMIS occurs.
> - anton
> --

Thanks by you remarks.
Could you please indicate which Forth you used for these usage values?
I'm curious to know what other types of words they could be and how they are implemented.
Note that DOCOL and SEMIS are only used for non-primitives words, for DOCODE, DOCON, DOVAR, etc as primitives, DOCOL and SEMIS are not used.
alvaro