• Open Menu Close Menu
  • Apple
  • Shopping Bag
  • Apple
  • Mac
  • iPad
  • iPhone
  • Watch
  • TV
  • Music
  • Support
  • Search apple.com
  • Shopping Bag

Lists

Open Menu Close Menu
  • Terms and Conditions
  • Lists hosted on this site
  • Email the Postmaster
  • Tips for posting to public mailing lists
Re: polling for bluetooth events
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: polling for bluetooth events


  • Subject: Re: polling for bluetooth events
  • From: Joseph Kelly <email@hidden>
  • Date: Mon, 10 Mar 2008 10:34:45 -0700

So basically you require a polling model -- occasionally you'll poll for any available input, process it, and then return to whatever it is you're doing. Unfortunately, the bluetooth api is dependent on the main thread run-loop (unlike HID which has a queueing mode), i.e. your main thread should spend its life handling events and otherwise remaining idle, and you would offload cpu intensive tasks to worker threads.

Have you considered handing off your bluetooth stuff to a separate process?

So basically, the process that needs to poll for events will spawn the bluetooth process (which creates a broadcast CFMessagePort), set up an event fifo, create a listener thread which creates a CFMessagePort to the bluetooth process. Then the bluetooth process connects to the device, sets up callbacks for data, then sits in the run loop. When data comes in from the wiimote to your bluetooth process' callbacks, it forwards the data over the message port to the listener thread, the listener thread stuffs the event into the event fifo. And then whenever your main process polls for events, it simply gets the lock on the event fifo, checks for available data, and unlocks.

It's not the cleanest solution, but I think it should get you a continuous flow of data from the device.

Joe K.

On Mar 7, 2008, at 6:08 PM, Hans-Christoph Steiner wrote:


Pd (aka Pure Data) has two processes that communicate via a network socket. There is a Tcl/Tk process that handles the GUI ('pd-gui') and the C process ('pd') that handles all of the digital signal processing. It is a realtime system, so it has its own cooperative scheduler. All processing needs to happen within the timeslice, or it will cause an interruption in the processing. So, for example, when I connect to the device, it causes everything to hang until the device is connected, and CFRunLoopStop() is called.


Since Tcl/Tk is based on Carbon, AFAIK, there is probably a CFRunLoop in it. But that's a separate process, so the 'pd' process isn't using any of that.

I wrote HID support for Pd, and I just polled the event queue, it works well. I was hoping for something like that for Bluetooth HID events. In particular, I am working on getting data from a Nintendo WiiRemote.

.hc

On Mar 7, 2008, at 8:34 PM, Joseph Kelly wrote:

Much of the user space Bluetooth API depends on run loops, i.e. for getting data to and from the kernel. It's surprising that your GUI framework itself does not rely on run loops. How does it respond to user input etc. ?

Unfortunately, I found that some Bluetooth calls only run reliably on the main thread, at least on Tiger and earlier.

Maybe you could explain your architecture a bit more. The model I've been using is I set up a callback for when data comes from the device, then I just return and wait; when the user does something which causes me to send commands to the device, and when I return from doing that (i.e. program flow returns to the event manager and the runloop code) I then get my data-in callbacks which I respond to.

Joe K.

On Mar 7, 2008, at 12:55 PM, Hans-Christoph Steiner wrote:


I am currently in the process of adding bluetooth HID support to Pure Data, a realtime, visual programming language for sound, video, etc. The GUI is all written in Tcl/Tk, so we are not handling a CFRunLoop at all.


I have gotten device searching and connection working by using CFRunLoopRun() then sticking CFRunLoopStop(CFRunLoopGetCurrent()) the end of the of the inquiry completion callback. I am trying to get the same thing working for getting events, but I haven't been able to.

Ideally, I would be able to poll for bluetooth events, is there a way to do that, outside of creating a separate thread to handle the event callbacks? Maybe something with CFRunLoopRunInMode()?

Here's the code in question:

http://pure-data.svn.sourceforge.net/viewvc/pure-data/trunk/ externals/io/wiiremote/

.hc


-------------------------------------------------------------------- --------


                                              http://at.or.at/hans/


_______________________________________________
Do not post admin requests to the list. They will be ignored.
Bluetooth-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
40mac.com


This email sent to email@hidden




---------------------------------------------------------------------- ------

All mankind is of one author, and is one volume; when one man dies, one chapter is not torn out of the book, but translated into a better language; and every chapter must be so translated.... -John Donne



_______________________________________________ Do not post admin requests to the list. They will be ignored. Bluetooth-dev mailing list (email@hidden) Help/Unsubscribe/Update your Subscription: This email sent to email@hidden
  • Follow-Ups:
    • Re: polling for bluetooth events
      • From: Hans-Christoph Steiner <email@hidden>
References: 
 >polling for bluetooth events (From: Hans-Christoph Steiner <email@hidden>)
 >Re: polling for bluetooth events (From: Joseph Kelly <email@hidden>)
 >Re: polling for bluetooth events (From: Hans-Christoph Steiner <email@hidden>)

  • Prev by Date: Re: polling for bluetooth events
  • Next by Date: Re: polling for bluetooth events
  • Previous by thread: Re: polling for bluetooth events
  • Next by thread: Re: polling for bluetooth events
  • Index(es):
    • Date
    • Thread