**** BEGIN LOGGING AT Sun May 25 02:59:59 2014 May 25 04:27:51 mranostay: What did you mean when you said, "You need to be sure one PRU transfers to shared memory and locks then copies to the local memory of PRU then kicks to userspace." May 25 06:08:47 vvu: I went past that error. May 25 06:10:13 The device went to ADK mode. May 25 06:13:03 I got an error in the mainPhase(). An I/o error it is. Since I'm able to get the device into ADK mode. This works. May 25 07:01:34 vvu: All errors resolved. May 25 07:02:47 Going through the comments on http://android.serverbox.ch/?p=262 helped. May 25 08:05:02 <_av500_> praveendath92: nice May 25 08:05:39 <_av500_> praveendath92: im flying to china for a week today May 25 08:05:44 <_av500_> ill be on irc sill May 25 08:05:45 <_av500_> still May 25 08:13:58 _av500_: Happy Journey :) . BTW, the ASM NEON memcpy crashes May 25 08:14:29 I get a segfault May 25 08:30:10 <_av500_> Abhishek_: hmm May 25 08:30:21 <_av500_> run it in gdb and see where May 25 08:30:54 I think it occurs in the memcpy only, if I replace the asm with the standard one, it works fine May 25 08:31:30 I am memcpying from a memmapped region (PRU mem) to a malloc region, would there be some problem with this? May 25 09:24:14 _av500_: Happy journey. I'm will be shifting this code into kernel module now. May 25 09:38:40 <_av500_> ok May 25 09:47:03 _av500_: Does it makes a difference to memcpy if I am copying from a mem-mapped region to a malloc region? May 25 09:53:04 _av500_: Using the v4si type (C SIMD) memcpy, the time is around 180us. May 25 09:54:06 I think the issue is not with an implementation of memcpy, it its due to access of DDR by both PRU and DDR, which might be causing stalls in accesses. May 25 09:54:13 *PRU and ARM Core May 25 11:23:07 _av500_: have a nice flight! May 25 11:53:04 <_av500_> thx May 25 15:33:16 panto : how do I give a platform driver a /dev entry? currently I managed to write a kernel module which does it with "register_chrdev" (the old way for char drivers if I'm correct). Would this also work for a platform driver? May 25 16:30:07 praveendath92 : hey! are you working with sysfs entries? May 25 16:30:56 karki: Hi. Not yet. May 25 16:31:10 I might have to. May 25 16:31:30 I have to, and I'm having some problems figuring my way out! May 25 16:31:40 btw what does your driver do as of now? May 25 16:31:57 * karki checks out bard-linux May 25 16:32:18 Right now the driver only have register and probing functions. May 25 16:32:35 I was testing a userspace code for a while. May 25 16:32:49 Now I will be porting that code into the driver. May 25 16:32:56 hhmm...... I think probe is the most complex part of my driver...... May 25 16:33:21 What all do you have to do in probe? May 25 16:33:32 set up the pru completely May 25 16:33:45 though it may be simpler that pantos code May 25 16:33:59 I don't have to handle vrings May 25 16:34:16 but have to create memory maps and sysfs entries. May 25 16:34:36 So I really don't know how messed up it'll be :/ May 25 16:34:53 you seem to be doing great though :) May 25 16:41:55 karki: client quit. May 25 16:41:59 Sorry. May 25 16:42:25 I think we should be minimizing the amount of work done in a probe function. May 25 16:42:51 karki: can you reuse the probe code from panto's implementation? May 25 16:43:11 Abhishek__ : about 50% I guess May 25 16:43:32 I will be doing sysfs, mmap which is currently present May 25 16:43:40 *currently not May 25 16:43:47 I see May 25 16:43:52 and I won't be doing vrings May 25 16:44:07 speed(mmap ) > speed(vring) May 25 16:44:12 vrings are there currently, right? May 25 16:44:16 yes May 25 16:45:23 the only thing I'm worried about is that it's quite hard to figure out what different parts of the current driver does. It's ~3K lines + I'm not good with platform drivers as of now! May 25 16:45:27 yep I'm aware of that May 25 16:45:42 ^ (vrings vs mmap) May 25 16:46:27 but with mmap you need a new signaling mechanism :/ May 25 16:46:40 so I'm trying to do that with sysfs May 25 16:46:46 oh May 25 16:47:10 you could refer to the UIO code for that I guess, it uses a similar function for interrupts IIRC May 25 16:47:27 I'm just wondering whether there is anyway to have a generic driver....... May 25 16:47:48 yes I had a look at UIO code :3 May 25 16:48:00 k May 25 16:49:12 Abhishek_ : but I'm not there yet. By this wednesday I want to get a simple platform driver with sysfs and mmap (interface only) ready. May 25 16:50:31 * karki is happy that his exams are over :) May 25 16:52:13 so, you live in B'lore too, or do you have to leave for home? May 25 16:55:25 I'm in B'lore. May 25 16:55:58 k May 25 16:56:01 I spend 3 hrs a day on travel :/ B'lore traffic May 25 16:56:55 * Abhishek_ remembers the times he was in Patna, IIT Patna was just 5 min away from my home ;) May 25 17:00:16 karki: So no uni acco? May 25 17:00:39 uni acco? May 25 17:00:52 university accomodation May 25 17:01:02 I didn't opt for it :( **** BEGIN LOGGING AT Sun May 25 23:47:05 2014 **** ENDING LOGGING AT Mon May 26 03:00:00 2014