-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 For the particularly daring...
On 12/25/2012 7:14 PM, Charles Steinkuehler wrote: > Of course, the proof is in the pudding, so keep an eye out on > Michael Haberler's git repo...I'll commit code there once it does > something useful. I have checked in some code to Michael's repository that does some (hopefully) useful tasks with the BeagleBone's PRU: http://git.mah.priv.at/gitweb/emc2-dev.git/shortlog/refs/heads/arm335x-hal-pru-module See the README.stepgen file for details: http://git.mah.priv.at/gitweb/emc2-dev.git/blob/786c7901faf01761359e0c7944f58407e2bfa10f:/configs/pru-examples/README.stepgen Executive Summary: With appropriate HAL hacking (and stuffing some configuration details directly into hal_pru.c), you can instantiate up to 7 PRU modules, each of which may currently be one of the following: Step/Dir generation Two "Traditional" PWMs Two 14-bit 1st order Delta-Sigma modulators ...and all in less than 512 bytes of PRU code (of the 8K available). TIMINGS: Currently the main loop is running on a 10uS timer, and if you build what's in the repository, apx. 1 uS is spent actually executing the code for one step/dir module and 6 PWM modules (12 PWM outputs). The timer is not yet 'interrupt' driven...there should be minimal latency jitter caused by the polling scheme, but jitter currently measures at about 100 nS. Fortunately, it does not appear to be additive over the long term. You can monitor the timings via a "busy" bit output on GPIO2_6 (P8, pin 45). If you change the configuration to 7 step/dir modules, code execution gets close to 2 uS. I have not yet analyzed worst-case code paths, and the TI documentation is pretty silent on cycle-times, but it feels like a 5 uS base thread (or faster) is possible with the generic code. WARNINGS: The PRU code is reasonably functional. The same cannot be said for the HAL modules that launch it. I crudely hacked Michael's hal_pru.c to initialize appropriate state data for the new stepgen.p PRU code for testing (thanks for the jump-start Michael!), but what is really needed is a mashup between the existing hal_pru.c (for PRU loading and interface), stepgen.c (to control the PRU stepgen base thread, which will behave mostly like the make_pulses() procedure), and something like the hostmot2 configurable HAL pin exporting for whatever PRU code modules a configuration happens to instantiate. And finally: NO WARRANTY EXPRESSED OR IMPLIED! This code *PROBABLY* won't destroy your BeagleBone, but no promises. :) I did have some issues with my 'bone, but popping the SD-Card and doing an fsck on another box fixed things right up (hint: don't manually set the I/O pin mux to enable *ALL* the PRU pins...leave the ones for the SD-Card interface alone!). There is a setup.sh file to keep you from making the same mistake...see the README.stepgen file for more of a HOWTO. Comments and criticism (especially of the PRU assembly code) will be cheerfully received! The PRU is a strange beast, and I'm still wrapping my head around programming it effectively... - -- Charles Steinkuehler char...@steinkuehler.net -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAlDdMpgACgkQLywbqEHdNFwxeACfcV13kv3vN19vySZeDtJ7B9xF Ps0AoJ9ZibJxs6Lx9QP3U4ohkzy7Mt73 =ZHcI -----END PGP SIGNATURE----- ------------------------------------------------------------------------------ Master HTML5, CSS3, ASP.NET, MVC, AJAX, Knockout.js, Web API and much more. Get web development skills now with LearnDevNow - 350+ hours of step-by-step video tutorials by Microsoft MVPs and experts. SALE $99.99 this month only -- learn more at: http://p.sf.net/sfu/learnmore_122812 _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
