highbit #52
Comments
|
I'm curious which floating point packs are freely available. Jos, can you point so something interesting ? I'd like to get a at least double precision library going for a project doing some astronomical toy (say, point antenna at moon) and most code for that does not really work with single precision. Also, am I right at suspecting that a 'proper' floating point implementation requires to change the interpreter to recognize FP numbers ? It seems feasible but seems like a fairly decent chunk of work, especially when FP speed is not of the essence (e.g. Moon angular speed relative to earth is not that fast). My other option, micropython (on esp32), is also handicapped in the precision department. |
|
Hi,
In the end I managed to port a simple high level floating point pack to
Punyforth.
There is no floating-point stack. It uses the normal stack.
2147483647 1+ is used as the highbit.
Not sure that is right for Punyforth.
The pack needs to be tested.
Interested?
Jos
From: klapauciusisgreat
Sent: Monday, May 25, 2020 12:00 AM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
I'm curious which floating point packs are freely available. Jos, can you
point so something interesting ?
I'd like to get a at least double precision library going for a project
doing some astronomical toy (say, point antenna at moon) and most code for
that does not really work with single precision.
Also, am I right at suspecting that a 'proper' floating point implementation
requires to change the interpreter to recognize FP numbers ? It seems
feasible but seems like a fairly decent chunk of work, especially when FP
speed is not of the essence (e.g. Moon angular speed relative to earth is
not that fast).
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
|
Yes, interested, as I didn't really find a high level FP library (I haven't
anything fort in a loong time).
Many thanks!
Max
…On Mon, May 25, 2020 at 2:42 AM Jos ***@***.***> wrote:
Hi,
In the end I managed to port a simple high level floating point pack to
Punyforth.
There is no floating-point stack. It uses the normal stack.
2147483647 1+ is used as the highbit.
Not sure that is right for Punyforth.
The pack needs to be tested.
Interested?
Jos
From: klapauciusisgreat
Sent: Monday, May 25, 2020 12:00 AM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
I'm curious which floating point packs are freely available. Jos, can you
point so something interesting ?
I'd like to get a at least double precision library going for a project
doing some astronomical toy (say, point antenna at moon) and most code for
that does not really work with single precision.
Also, am I right at suspecting that a 'proper' floating point
implementation
requires to change the interpreter to recognize FP numbers ? It seems
feasible but seems like a fairly decent chunk of work, especially when FP
speed is not of the essence (e.g. Moon angular speed relative to earth is
not that fast).
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#52 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AGMK7CPUZXRND3DV7OC3EMLRTI4RJANCNFSM4IOTGKEQ>
.
|
|
Hi,
See zenfloat.zip at the end of https://sites.google.com/site/win324th/smart-home
for a download.
Jos
From: Klapaucius
Sent: Monday, May 25, 2020 5:54 PM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
Yes, interested, as I didn't really find a high level FP library (I haven't
anything fort in a loong time).
Many thanks!
Max
On Mon, May 25, 2020 at 2:42 AM Jos ***@***.***> wrote:
Hi,
In the end I managed to port a simple high level floating point pack to
Punyforth.
There is no floating-point stack. It uses the normal stack.
2147483647 1+ is used as the highbit.
Not sure that is right for Punyforth.
The pack needs to be tested.
Interested?
Jos
From: klapauciusisgreat
Sent: Monday, May 25, 2020 12:00 AM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
I'm curious which floating point packs are freely available. Jos, can you
point so something interesting ?
I'd like to get a at least double precision library going for a project
doing some astronomical toy (say, point antenna at moon) and most code for
that does not really work with single precision.
Also, am I right at suspecting that a 'proper' floating point
implementation
requires to change the interpreter to recognize FP numbers ? It seems
feasible but seems like a fairly decent chunk of work, especially when FP
speed is not of the essence (e.g. Moon angular speed relative to earth is
not that fast).
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#52 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AGMK7CPUZXRND3DV7OC3EMLRTI4RJANCNFSM4IOTGKEQ>
.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
|
Very cool, thanks! It's always amazing how much you can do with so little
code in forth (though APL seems equally compact).
This won't be enough for me since for pointing at the moon I need a bunch
of trig functions. I'll consider adding floating point to punyforth
natively, but I first need to get it to work on esp32 devices at all. The
code is a bit gnarly - both on the esp-idf sdk and punyforth side, but it
seems doable.
I also found some hints on how to make micropython double precision, I
will try that as well. It seems like a simpler way forward, and most esp32
devices have no problem running substantial python programs.
Anyway, thanks again (also for the winforth pages -- nice!)
Max
…On Tue, May 26, 2020 at 2:42 AM Jos ***@***.***> wrote:
Hi,
See zenfloat.zip at the end of
https://sites.google.com/site/win324th/smart-home
for a download.
Jos
From: Klapaucius
Sent: Monday, May 25, 2020 5:54 PM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
Yes, interested, as I didn't really find a high level FP library (I haven't
anything fort in a loong time).
Many thanks!
Max
On Mon, May 25, 2020 at 2:42 AM Jos ***@***.***> wrote:
> Hi,
>
> In the end I managed to port a simple high level floating point pack to
> Punyforth.
> There is no floating-point stack. It uses the normal stack.
> 2147483647 1+ is used as the highbit.
> Not sure that is right for Punyforth.
> The pack needs to be tested.
>
> Interested?
> Jos
>
>
> From: klapauciusisgreat
> Sent: Monday, May 25, 2020 12:00 AM
> To: zeroflag/punyforth
> Cc: Jos ; Author
> Subject: Re: [zeroflag/punyforth] highbit (#52)
>
>
> I'm curious which floating point packs are freely available. Jos, can you
> point so something interesting ?
>
> I'd like to get a at least double precision library going for a project
> doing some astronomical toy (say, point antenna at moon) and most code
for
> that does not really work with single precision.
>
> Also, am I right at suspecting that a 'proper' floating point
> implementation
> requires to change the interpreter to recognize FP numbers ? It seems
> feasible but seems like a fairly decent chunk of work, especially when FP
> speed is not of the essence (e.g. Moon angular speed relative to earth is
> not that fast).
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub, or unsubscribe.
>
> —
> You are receiving this because you are subscribed to this thread.
> Reply to this email directly, view it on GitHub
> <#52 (comment)
>,
> or unsubscribe
> <
https://github.com/notifications/unsubscribe-auth/AGMK7CPUZXRND3DV7OC3EMLRTI4RJANCNFSM4IOTGKEQ
>
> .
>
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#52 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AGMK7CIRVO2O2YBRP2KQOSDRTOFG3ANCNFSM4IOTGKEQ>
.
|
|
Hi,
Please inform us when punyforth is working on a esp32.
Jos
From: Klapaucius
Sent: Wednesday, May 27, 2020 6:20 PM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
Very cool, thanks! It's always amazing how much you can do with so little
code in forth (though APL seems equally compact).
This won't be enough for me since for pointing at the moon I need a bunch
of trig functions. I'll consider adding floating point to punyforth
natively, but I first need to get it to work on esp32 devices at all. The
code is a bit gnarly - both on the esp-idf sdk and punyforth side, but it
seems doable.
I also found some hints on how to make micropython double precision, I
will try that as well. It seems like a simpler way forward, and most esp32
devices have no problem running substantial python programs.
Anyway, thanks again (also for the winforth pages -- nice!)
Max
On Tue, May 26, 2020 at 2:42 AM Jos ***@***.***> wrote:
Hi,
See zenfloat.zip at the end of
https://sites.google.com/site/win324th/smart-home
for a download.
Jos
From: Klapaucius
Sent: Monday, May 25, 2020 5:54 PM
To: zeroflag/punyforth
Cc: Jos ; Author
Subject: Re: [zeroflag/punyforth] highbit (#52)
Yes, interested, as I didn't really find a high level FP library (I
haven't
anything fort in a loong time).
Many thanks!
Max
On Mon, May 25, 2020 at 2:42 AM Jos ***@***.***> wrote:
> Hi,
>
> In the end I managed to port a simple high level floating point pack to
> Punyforth.
> There is no floating-point stack. It uses the normal stack.
> 2147483647 1+ is used as the highbit.
> Not sure that is right for Punyforth.
> The pack needs to be tested.
>
> Interested?
> Jos
>
>
> From: klapauciusisgreat
> Sent: Monday, May 25, 2020 12:00 AM
> To: zeroflag/punyforth
> Cc: Jos ; Author
> Subject: Re: [zeroflag/punyforth] highbit (#52)
>
>
> I'm curious which floating point packs are freely available. Jos, can
> you
> point so something interesting ?
>
> I'd like to get a at least double precision library going for a project
> doing some astronomical toy (say, point antenna at moon) and most code
for
> that does not really work with single precision.
>
> Also, am I right at suspecting that a 'proper' floating point
> implementation
> requires to change the interpreter to recognize FP numbers ? It seems
> feasible but seems like a fairly decent chunk of work, especially when
> FP
> speed is not of the essence (e.g. Moon angular speed relative to earth
> is
> not that fast).
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub, or unsubscribe.
>
> —
> You are receiving this because you are subscribed to this thread.
> Reply to this email directly, view it on GitHub
> <#52 (comment)
>,
> or unsubscribe
> <
https://github.com/notifications/unsubscribe-auth/AGMK7CPUZXRND3DV7OC3EMLRTI4RJANCNFSM4IOTGKEQ
>
> .
>
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#52 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AGMK7CIRVO2O2YBRP2KQOSDRTOFG3ANCNFSM4IOTGKEQ>
.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi,
At I am still quite happy with your Punyforth. Great work!
There is just 1 question for a primitive floating-point pack.
When highbit is defined as:
-1 1 rshift invert constant: highbit
than highbit returns 214748364
Other 32 bits systems return 2147483648
How should highbit be defined?
Thanks in advance.
Despite that problem the pack seem to work as you can see:
(stack) 10 s>f 3 s>f f/ f.
333333333e-8(stack)
10 s>f 6 s>f f/ f.
1666666666e-9(stack)
Kind regards, Jos
The text was updated successfully, but these errors were encountered: