Shop OBEX P1 Docs P2 Docs Learn Events
Make a Faster I/O (Advanced Programmers Only) — Parallax Forums

Make a Faster I/O (Advanced Programmers Only)

HumanoidoHumanoido Posts: 5,770
edited 2010-09-17 23:56 in Propeller 1
Who can make a faster I/O?

As pointed out by Ali and others, the Propeller's I/O speed is limited. Recent posts show ideas to overlay and make use of high speed Counters.

Who can create a faster I/O example? Make it simple, one pin.

Thanks in advance.
Humanoido

http://forums.parallax.com/showthread.php?t=125627

Comments

  • Cluso99Cluso99 Posts: 18,069
    edited 2010-09-16 09:18
    I have a datalogger that uses 4 cogs which are interleaved to sample all 32 pins on each clock cycle. IIRC it performs 1700 samples before it fills cog memory and pushes the data to the serial port. Mine is not the only datalogger that interleaves like this.
  • HumanoidoHumanoido Posts: 5,770
    edited 2010-09-16 09:46
    Cluso99 wrote: »
    I have a datalogger that uses 4 cogs which are interleaved to sample all 32 pins on each clock cycle. IIRC it performs 1700 samples before it fills cog memory and pushes the data to the serial port. Mine is not the only datalogger that interleaves like this.
    Cluso, I believe this method is very useful. Can you provide more information about it? I was looking at the results from Beau, Heater, and Kuroneko on a slightly larger cluster prop running a pure speed algorithm and it comes out to a max of .32 Teraflop using all the cogs with the first stage of overclocking (6.25). Working together on this, we can come up with something profound.

    Humanoido
  • User NameUser Name Posts: 1,451
    edited 2010-09-16 14:18
    Why would you create a thread that excludes yourself? What realities are you running away from?
  • Cluso99Cluso99 Posts: 18,069
    edited 2010-09-16 15:49
    I posted the datalogger code on the OBEX. It is quite well documented IIRC.
  • HumanoidoHumanoido Posts: 5,770
    edited 2010-09-17 01:07
    Cluso99 wrote: »
    I posted the datalogger code on the OBEX. It is quite well documented IIRC.
    Which one, any of these?

    http://obex.parallax.com/objects/search/?q=datalogger&csrfmiddlewaretoken=0a03997685b76fdcfa02de8a0bc8b0e1

    I.K
  • Heater.Heater. Posts: 21,230
    edited 2010-09-17 01:38
    The data logger is great but it does not answer the question of this thread.
    Who can create a faster I/O example? Make it simple, one pin.
    Also it uses 4 COGs which means when applied to any multi-prop compute system it has halved your available compute power.

    Somewhere on this forum is described a fast serial connection between Props that is 7MBits per second (or so) if I remember correctly. Can't find it now. If that were done in half duplex one would have about 3Mbits/sec each way, only one pin and one COG used.

    Ah here it is, the BeauNet http://forums.parallax.com/showthread.php?t=99222&highlight=High+Speed+Prop-to-Prop+Serial+Communication

    Seems to be up to 14.5MBits/second now :)
  • Cluso99Cluso99 Posts: 18,069
    edited 2010-09-17 02:07
    Here is the OBEX link http://obex.parallax.com/objects/572/
  • HumanoidoHumanoido Posts: 5,770
    edited 2010-09-17 03:13
    NEW 12/12/2007
    14.5 Meg Baud Upgrade

    does that link work for you???
  • Heater.Heater. Posts: 21,230
    edited 2010-09-17 03:24
    No but you will find it on page 2 of the thread I linked to (which does work)

    14.5 MBaud uses 2 Cogs and 4 pins. 2 Tx and 2 Rx.

    I really have to find time to try that out one day.
  • evanhevanh Posts: 16,161
    edited 2010-09-17 19:52
    Hehe, interpretations of what is meant by "I/O".

    - The average IT geek thinks files.

    - The average CS major thinks data bus.

    - The average EE major thinks sampling.
  • Heater.Heater. Posts: 21,230
    edited 2010-09-17 23:56
    And I think:

    "I/O, I/O, it's in and out we go..."

    The IT geek, CS major and EE major all find out it all comes down to the same thing on the Prop. How fast can we wiggle and read those pins using software.
Sign In or Register to comment.