[Bf-committers] An easy to use damped spline

Sat Dec 27 11:39:11 CET 2014

Ugh, never mind, floating point splines aren't any trouble :)

https://gitlab.com/ideasman42/fred-curve/blob/float_spline/pspline.c


On Sat, Dec 27, 2014 at 9:23 PM, Campbell Barton <ideasman42 at gmail.com> wrote:
> Heres an initial port of the PSpline.pas to C,
> (single C file, prints 2d curve to stdout).
>
> https://gitlab.com/ideasman42/fred-curve/blob/HEAD/pspline.c
>
> This is the n-dimensional variant, but for the test it only prints 2d values.
>
> My main concern with this method of calculating splines is it relies
> on integer values and bit-shifting. For Blender we would want it to
> calculate floating point values, I didnt investigate yet how well this
> method could be used with floats, would be interesting to check on
> next.
>
> On Mon, Sep 29, 2014 at 3:45 AM, Fred <f.harthoorn at ziggo.nl> wrote:
>> Hi Campbell Barton,
>>
>> The amount of interpolation dots is 2^i, at the start of the demo  i =
>> 4    (16 dots)
>> if you change i = 7 (the limit in the demo) than there are 128
>> interpolation dots.
>>
>> if I had set the limit in the demo to 10 then 1024 interpolation dots
>> was possible.
>>
>> If the spline is used in a 3 dimensional space,
>> to compute an interpolated dot position takes:  dim * 10  clock cycles =
>> 30 clock cycles.
>> I hope that it is fast enough.
>>
>> If you like to have a curve to the last plotted dot you have to put two
>> dots on each other at the end.
>> You will see that it is symmetrical again.
>>
>> Please change the number to 7 in the demo and you see a curved line in
>> stead of dots.
>>
>> Much more is possible with this spline because the first, second and
>> third derivative (parameterized ) of the curve
>> is available in real time.
>>
>> I think that much more can be invented with this spline technique.
>> I think that with some research tangent handles can be constructed as
>> well (if you like them?).
>> What I did in the demo was the most simple way of using it.
>>
>> The essence is: a 3th degree curve in two (x,y) or more directions is
>> differentiated three times, and then again integrated 3 times.
>> With floating point this is impossible , because it is not exact, the
>> curve runs away.
>> But with integer computing it is exact and the spline exact a 3th degree
>> curve.
>>
>> It can also be used in ten or more directions, but more then 3
>> dimensions (x,y,z) is not  practical.
>>
>> Fred.
>> --------------------------------------
>> Fred schreef op 28-9-2014 om 18:30:
>>> That is in the demo the case, because I programmed that  the first plot
>>> point is actually two points.
>>> the spline uses 4 points p0, p1, p2, p3  and the interpolation is
>>> between p1 and p2.
>>>
>>> Campbell Barton schreef op 28-9-2014 om 17:42:
>>>> Checked the test application, and not sure if this is intended but a
>>>> simple triangle has uneven distribution.
>>>>
>>>> see: http://www.graphicall.org/ftp/ideasman42/fred_curve.png
>>>>
>>>> On Mon, Sep 29, 2014 at 1:32 AM, Dan McGrath <danmcgrath.ca at gmail.com> wrote:
>>>>> Cool, got them. Will take a look later. o/
>>>>>
>>>>> Dan
>>>>>
>>>>> On Sun, Sep 28, 2014 at 11:28 AM, Fred <f.harthoorn at ziggo.nl> wrote:
>>>>>
>>>>>> Ok, It is uploaded to the directory incoming
>>>>>>
>>>>>> Dan schreef op 28-9-2014 om 17:15:
>>>>>>> You could always upload it to our anonymous ftp via download.blender.org
>>>>>>> (or ftp.blender.org) and put it into the incoming directory (use passive
>>>>>>> mode). From there was can fetch the files via
>>>>>>> http://download.blender.org/ftp/incoming/
>>>>>>>
>>>>>>> Dan
>>>>>>>
>>>>>>> On Sun, Sep 28, 2014 at 11:05 AM, Fred <f.harthoorn at ziggo.nl> wrote:
>>>>>>>
>>>>>>>> Campbell Barton
>>>>>>>>
>>>>>>>> I have tried to send it but it is rejected. So send me an addres for
>>>>>>>> upload.
>>>>>>>>
>>>>>>>> fred
>>>>>>>>
>>>>>>>>
>>>>>>>> Campbell Barton schreef op 28-9-2014 om 16:51:
>>>>>>>>> Could the project files be posted to a public URL?, Id like to be able
>>>>>>>>> to look into this. but probably wont have time over the coming weeks.
>>>>>>>>>
>>>>>>>>> The main source file compiles for me with minor edits on
>>>>>>>>> Linux/FreePascal, but the demo relies on `Window` unit, which I assume
>>>>>>>>> is provided by delphi/windows.
>>>>>>>>>
>>>>>>>>> On Mon, Sep 29, 2014 at 12:36 AM, Dan McGrath <danmcgrath.ca at gmail.com
>>>>>>>> wrote:
>>>>>>>>>> Hey,
>>>>>>>>>>
>>>>>>>>>> Just save out the delphi project and zip it up and email it to me
>>>>>> direct
>>>>>>>>>> via email I guess (the mailing list wouldn't allow such a large
>>>>>>>>>> attachement). I was attempting to compile it on linux, so the compiled
>>>>>>>>>> dll's wouldn't be of much use to me unfortunately.
>>>>>>>>>>
>>>>>>>>>> Thanks!
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Dan
>>>>>>>>>>
>>>>>>>>>> On Sun, Sep 28, 2014 at 10:06 AM, Fred <f.harthoorn at ziggo.nl> wrote:
>>>>>>>>>>
>>>>>>>>>>> Hi Dan,
>>>>>>>>>>>
>>>>>>>>>>> Of course I can  send you the project.
>>>>>>>>>>> I have made two versions:
>>>>>>>>>>> 1) The spline is embedded in de demo:
>>>>>>>>>>> 2) The spline is an external .dll, de demo calls PSpline.dll
>>>>>>>>>>>
>>>>>>>>>>> Which one do you prefer?
>>>>>>>>>>>
>>>>>>>>>>> Fred.
>>>>>>>>>>> ----------------
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Dan McGrath schreef op 28-9-2014 om 15:42:
>>>>>>>>>>>> So is/was Turbo Pascal (limited in features; eg: objects but no
>>>>>>>> classes)
>>>>>>>>>>>> and FreePascal+Lazarus (supports almost everything Delphi does
>>>>>> still?
>>>>>>>>>>>> http://wiki.freepascal.org/Lazarus_For_Delphi_Users).
>>>>>>>>>>>>
>>>>>>>>>>>> Anyway, just thought it was neat to see code in pascal (an elegant
>>>>>>>>>>>> language, imho). Curious though, why not just give the whole delphi
>>>>>>>>>>> project
>>>>>>>>>>>> and resource files so your forms can be built etc.? The code seems
>>>>>> to
>>>>>>>>>>>> compile with some modifications here, but without the {$R *.DFM}
>>>>>>>> (*.lfm
>>>>>>>>>>> in
>>>>>>>>>>>> fp), one would have to manually rebuild the form :(
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> On Sun, Sep 28, 2014 at 8:55 AM, Fred <f.harthoorn at ziggo.nl> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> It's true, but Delphi is object oriented.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Dan McGrath schreef op 28-9-2014 om 14:53:
>>>>>>>>>>>>>> Standard Pascal, Turbo Pascal, Freepascal, Delphi; same language,
>>>>>>>>>>>>> different
>>>>>>>>>>>>>> dialect :)
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On Sun, Sep 28, 2014 at 8:43 AM, Fred <f.harthoorn at ziggo.nl>
>>>>>> wrote:
>>>>>>>>>>>>>>> Actually it is Delphi
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Dan McGrath schreef op 28-9-2014 om 14:35:
>>>>>>>>>>>>>>>> Pascal, with inline assembler to boot! \o/ :)
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On Sun, Sep 28, 2014 at 6:46 AM, Ton Roosendaal <
>>>>>> ton at blender.org>
>>>>>>>>>>>>> wrote:
>>>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Fred told me he's not interested in getting involved, or
>>>>>> setting
>>>>>>>> up
>>>>>>>>>>> a
>>>>>>>>>>>>>>>>> project for his work.
>>>>>>>>>>>>>>>>> He's retired now and just thought we might be interested in a
>>>>>>>>>>> research
>>>>>>>>>>>>>>>>> topic he did.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> I've copied files he sent to me here;
>>>>>>>>>>>>>>>>> http://download.blender.org/ftp/ton/
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> It's a Dutch pdf and two pascal files.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> -Ton-
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> --------------------------------------------------------
>>>>>>>>>>>>>>>>> Ton Roosendaal  -  ton at blender.org   -   www.blender.org
>>>>>>>>>>>>>>>>> Chairman Blender Foundation - Producer Blender Institute
>>>>>>>>>>>>>>>>> Entrepotdok 57A  -  1018AD Amsterdam  -  The Netherlands
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 28 Sep, 2014, at 9:28, Aditia A. Pratama wrote:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> HI Fred,
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Can you record a screencast/video demo to show how it works?
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Regards,
>>>>>>>>>>>>>>>>>> On Sep 28, 2014 2:14 PM, "Fred" <f.harthoorn at ziggo.nl> wrote:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> I forgot to say, I have a two dimensional demo but it is
>>>>>>>> working
>>>>>>>>>>> on
>>>>>>>>>>>>> a
>>>>>>>>>>>>>>>>>>> windows computer.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Jacob Merrill schreef op 28-9-2014 om 02:07:
>>>>>>>>>>>>>>>>>>>> would this make it so you can record F-curves over time
>>>>>>>> better?
>>>>>>>>>>>>>>>>>>>> right now recording in the engine gets slower the longer it
>>>>>>>> runs,
>>>>>>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>>
>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>>
>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>>
>>>>>>>>>>> _______________________________________________
>>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>>>>
>>>>>>>>>> _______________________________________________
>>>>>>>>>> Bf-committers mailing list
>>>>>>>>>> Bf-committers at blender.org
>>>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>> _______________________________________________
>>>>>>>> Bf-committers mailing list
>>>>>>>> Bf-committers at blender.org
>>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Bf-committers mailing list
>>>>>>> Bf-committers at blender.org
>>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>>
>>>>>> _______________________________________________
>>>>>> Bf-committers mailing list
>>>>>> Bf-committers at blender.org
>>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>>>
>>>>> _______________________________________________
>>>>> Bf-committers mailing list
>>>>> Bf-committers at blender.org
>>>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>>
>>> _______________________________________________
>>> Bf-committers mailing list
>>> Bf-committers at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-committers
>>>
>>
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-committers
>
>
>
> --
> - Campbell


-- 
- Campbell