PNut/Spin2 Latest Version (v34z - Lots of DEBUG display stuff)

1282930313234»

Comments

  • Thank you, Chip!
  • JonnyMac wrote: »
    This may not be compatible with FastSpin (which I don't use), but it does seem to work in standard Spin. My MS_001 constant is configured for 200MHz which I have standardized on.

    To make it clear: I love interrupts, even if they can be seen to behave a little "insidiously", insisting on "infiltrating" wherever they can, whenever one is not expecting them to be even "admissible".

    That said, perhaps there is another way to encapsulate such kind of code, without ever having to take care about any previous condition of interrupts (being enabled or not/occuring or not) and, at the same time, sparing one instruction:

    pub get_ms() : ms | lo, hi
    
    '' Return milliseconds after reset (32 bits).
    '' -- system counter is fixed; cannot be changed by user
    
      org
                    rep       #5,#1                  ' executes once; protects code block from being interrupted
                    getct     hi                            wc      ' get cnt (now)
                    getct     lo
    '                stalli                                          ' stall interrupts to protect cordic operation
                    setq      hi                                    ' divide cnt by ticks/ms
                    qdiv      lo, ##MS_001
                    getqx     ms
    '                allowi                                          ' allow interrupts
      end
    

    Hope it could help :smile:

    Henrique
  • Lol, that's instantly made two instructions without a purpose now.

  • JonnyMacJonnyMac Posts: 7,008
    edited 2020-09-22 - 01:06:13
    That said, perhaps there is another way to encapsulate such kind of code, without ever having to take care about any previous condition of interrupts (being enabled or not/occuring or not) and, at the same time, sparing one instruction:
    I was following what Chip does in the Spin interpreter for the getsec() instruction.
  • To make GETMS () work, I took the remainder from the GETSEC () division, did a QFRAC with CLKFREQ, multiplied that by 1000 and took the upper long of the product. This gives GETMS () the same lifespan as GETSEC (), which is $1_0000_0000 seconds, or 136.19 years. In that time, GETMS () will roll over 1,000 times before both GETSEC () and GETMS () will return $FFFF_FFFF and some other value, terminally.
  • Roy ElthamRoy Eltham Posts: 2,920
    edited 2020-09-22 - 05:37:02
    Does that lose resolution of getms()? You want millisecond resolution for getms() so it should roll over much faster.

    or are you saying that you have to do getsec() and then getms() after to get the ms portion? That sounds terribad also.
  • cgraceycgracey Posts: 13,079
    edited 2020-09-22 - 05:48:07
    Roy Eltham wrote: »
    Does that lose resolution of getms()? You want millisecond resolution for getms() so it should roll over much faster.

    or are you saying that you have to do getsec() and then getms() after to get the ms portion? That sounds terribad also.

    After computing seconds, which is the 64-bit counter divided by CLKFREQ, you have remainder between 0 and CLKFREQ-1. There is plenty of resolution to determine milliseconds from that remainder. And the milliseconds are always congruous to the seconds. I thought about many ways in which to compute milliseconds, and this was the only one that was practical.

    You just do a GETMS () instruction.
  • This sounds good. So do we have both a GETMS() and a GETSEC() method now, is that how it works? I had thought you needed to replace GETSEC() to make GETMS() fit?
  • TonyB_TonyB_ Posts: 1,524
    edited 2020-09-22 - 13:19:37
    Yanomani wrote: »
    [
    To make it clear: I love interrupts, even if they can be seen to behave a little "insidiously", insisting on "infiltrating" wherever they can, whenever one is not expecting them to be even "admissible".

    That said, perhaps there is another way to encapsulate such kind of code, without ever having to take care about any previous condition of interrupts (being enabled or not/occuring or not) and, at the same time, sparing one instruction:

    pub get_ms() : ms | lo, hi
    
    '' Return milliseconds after reset (32 bits).
    '' -- system counter is fixed; cannot be changed by user
    
      org
                    rep       #5,#1                  ' executes once; protects code block from being interrupted
                    getct     hi                            wc      ' get cnt (now)
                    getct     lo
    '                stalli                                          ' stall interrupts to protect cordic operation
                    setq      hi                                    ' divide cnt by ticks/ms
                    qdiv      lo, ##MS_001
                    getqx     ms
    '                allowi                                          ' allow interrupts
      end
    

    Hope it could help :smile:

    Henrique

    Excellent! Disabling Blocking interrupts with REP should go in the Tips & Tricks thread, wherever that is now.
  • rogloh wrote: »
    This sounds good. So do we have both a GETMS() and a GETSEC() method now, is that how it works? I had thought you needed to replace GETSEC() to make GETMS() fit?

    Yes, we have both. I hope to get a new version out in the next few days.
  • Hi evanh...

    I know that you know about it, but, in case someone just feels confused by :smile:

    ... and thanks TonyB, acknowledgeing your suggestion:

    https://forums.parallax.com/discussion/comment/1507194/#Comment_1507194
  • (I'm sorry if this is off topic... but I know that Chip reads this thread!)

    @Chip
    Do have a midi object that I can use? I saw your awesome presentation on audio processing with pitchbend, chourus and midi on YouTube. I don't want to reinvent the wheel if there is something I can just drop in and use out of the box.
  • Ahle2 wrote: »
    (I'm sorry if this is off topic... but I know that Chip reads this thread!)

    @Chip
    Do have a midi object that I can use? I saw your awesome presentation on audio processing with pitchbend, chourus and midi on YouTube. I don't want to reinvent the wheel if there is something I can just drop in and use out of the box.

    All I have is in the zip file at the top of this thread. To run DEBUG_MIDI, you'll need to get MIDI into the P2. I used this:

    http://ubld.it/products/midi-breakout-board/
  • JonnyMacJonnyMac Posts: 7,008
    edited 2020-09-24 - 16:52:33
    Ahle2 wrote: »
    @Chip
    Do have a midi object that I can use? I saw your awesome presentation on audio processing with pitchbend, chourus and midi on YouTube. I don't want to reinvent the wheel if there is something I can just drop in and use out of the box.

    The DEBUG_MIDI program doesn't actually process the MIDI data; it simply captures the serial input from the MIDI device and passes it to the MIDI DEBUG window which handles the Note On and Note Off messages.

    I am starting on a standard MIDI object. What features are you looking for? I'm trying to write an object that can accommodate all legal MIDI 1.0 messages, including the system level messages.

  • @Chip
    Thanks.. I will look in the archive for any goodies! :smile:

    @JonnyMac
    Great to hear that you have started implementing a MIDI object, it will be plenty useful for many things I do. For now I will just use the basic features of the MIDI 1.0 standard. Note on/off, control change, patch change, ptich bend etc.
  • cgraceycgracey Posts: 13,079
    edited 2020-10-12 - 13:05:56
    I got interested in anti-aliasing graphics for the DEBUG displays, because pixels are so coarse. I worked all night and made an anti-aliasing line-draw routine in Delphi. I need to fix it up so that it properly positions at 1/$10000 screen pixels for start and end points.

    antialiasing.png
    344 x 351 - 7K
  • That's great, Chip! Looking forward to it.
  • Looks like the midnight oil paid off!
  • Publison wrote: »
    Looks like the midnight oil paid off!

    I was thinking about this yesterday, looking at various graphics packages I could use, but everything was overwrought. I finally just figured I could probably make it, myself. I need to add clipping to it, yet.
  • Yes, Looks great!
  • I've almost got the anti-aliasing done. The difference in quality is huge.

    scope_antialiased.png
    588 x 172 - 6K
    274 x 150 - 272K
  • Looks very nice. But for the scope waveform I wonder if the anti aliasing would probably mask small steps, slow changes or dithering. I'd prefer seeing as many details as possible even if it doesn't look nice. For all other graphics like the hub/cog wheel it is definitively an improvement.
  • Chip, to @ManAtWork's point, maybe an option to turn off antialiasing on a per-Debug basis?
  • JonnyMacJonnyMac Posts: 7,008
    edited 2020-10-13 - 14:55:19
    <Party pooper mode>These are not lab grade 'scopes, and quibbling over features is only delaying the release of Propeller Tool with DEBUG features integrated.</PPM>
  • ManAtWork wrote: »
    Looks very nice. But for the scope waveform I wonder if the anti aliasing would probably mask small steps, slow changes or dithering. I'd prefer seeing as many details as possible even if it doesn't look nice. For all other graphics like the hub/cog wheel it is definitively an improvement.

    I used to think this, too, but what it really does is increase the resolution by 256x.

    I think what you are asking for is a mode where each sample is shown as a vertically-positioned horizontal line. We can do that better with anti-aliasing.
  • ElectrodudeElectrodude Posts: 1,434
    edited 2020-10-15 - 20:57:19
    cgracey wrote: »
    Publison wrote: »
    Looks like the midnight oil paid off!

    I was thinking about this yesterday, looking at various graphics packages I could use, but everything was overwrought. I finally just figured I could probably make it, myself. I need to add clipping to it, yet.

    Take another look at OpenGL with GLFW. OpenGL has a lot of complications you can probably ignore since you aren't writing a game with fancy graphics.

    This introduction looks like it might be good for you: https://en.wikibooks.org/wiki/OpenGL_Programming/Scientific_OpenGL_Tutorial_01. You'll probably be happier ignoring all the stuff there about GLUT and using GLFW instead, since GLUT forces you to use its own main loop while GLFW lets you have your own; you can find a GLFW example which draws a spinning colored triangle here, with an explanation here; start with this and replace the drawing stuff and shaders with the ones in the Scientific OpenGL tutorial.

    There are Delphi bindings for GLFW and OpenGL here: https://github.com/neslib/DelphiGlfw. Don't use the old glNewList() calls their examples use; do it as in the Scientific OpenGL Tutorial with gl{Gen,Bind,Map}Buffer-type functions. You should be able to get OpenGL to directly process your plot data without having to convert it to some other format first - this is what I do in my plotting software.

    I should be able to help you with this, since I have experience with drawing graphs in OpenGL. I hope I can publish my graphing software soon, together with some interesting Propeller examples.
  • Electrodude, thanks for the pointers. I will look into that after I get a few more things out of the way. I've almost got the anti-aliasing done, but I will need to get some other compiler work finished before I can get back into this graphic stuff, which is really fascinating.
  • cgraceycgracey Posts: 13,079
    edited 2020-10-21 - 11:47:54
    Last week I had to sit for 5 hours, so I brought a clipboard with me and worked out how to do 2D anti-aliasing. I'm now drawing anti-aliased shapes faster than Windows' built-in routines can draw aliased shapes.

    This is what PLOT output looks like now:

    PLOT_anti_aliasing.png

    I found a way to do it that is very simple and deterministic. I want to make a PLOT animation to demonstrate the technique. A kindergartner could understand the principle, but it was the hardest thing I've worked out in a few years. If I hadn't been stuck sitting, I might not have gotten in done.

    Now I'm working on a line-draw that has begin- and end-points in 256ths of a pixel. This is needed for the scope-type displays. I kind of had this working, at first, but it used whole pixel positions for start and end and the the endpoint pixels were messy. This new one is going to be as good as possible.

    606 x 679 - 22K
Sign In or Register to comment.