**** BEGIN LOGGING AT Fri Jun 12 02:59:59 2015
Jun 12 09:07:29 <Abhishek_> shubhangi: Also consider how are interrupts supposed to be passed to userspace / kernel drivers seeking to interact with the PRU
Jun 12 13:23:46 <Abhishek_> shubhangi: Had responded, check the logs
Jun 12 13:32:13 <shubhangi> Abhishek_: the response about interrupts to userspace ?
Jun 12 13:32:34 <shubhangi> Abhishek_: or am i missing something else.
Jun 12 14:16:42 <shubhangi> Abhishek_: ping ^
Jun 12 14:43:37 <Abhishek_> shubhangi: yes
Jun 12 14:45:07 <shubhangi> Abhishek_: beaglelogic is using 4mb buffers because allocating large contiguous memory is not easy. right ?
Jun 12 14:46:33 <Abhishek_> shubhangi: Yes, but the other reason is that at the highest possible sample rate, a buffer size any less would cause a lot of interrupts to occur.
Jun 12 14:48:53 <Abhishek_> at 100 MSa/s , 4 MB is like ~42ms
Jun 12 14:49:45 <shubhangi> Abhishek_: If i can manage a single 32M buffer using CMA or some other techinque would that be enough
Jun 12 14:50:22 <shubhangi> Abhishek_: or smaller circularly linked buffers should be the way forward. this way i can have a lot of memory
Jun 12 14:50:50 <shubhangi> Abhishek_: single large buffer can make use of linux fifo library. very well written
Jun 12 14:50:51 <Abhishek_> BeagleLogic uses (4 MB) circularly linked buffers
Jun 12 14:51:10 <shubhangi> yes. that i know
Jun 12 14:52:45 <Abhishek_> Ask panto for his thoughts on the use of CMA ( I do not recommend it though, would have used it for BeagleLogic otherwise ) , or carry on with whatever largest size buffer you are able to allocate.
Jun 12 14:53:08 <shubhangi> theres so much out there on memory managment :O . just need to make sure i get the best way out
Jun 12 14:53:25 <Abhishek_> memory management is but one part
Jun 12 14:54:42 <shubhangi> i'll get the interrupts to userspace. however i see people are against handling them in userspace. i should just just probable acknowdge them in userspace and handle it by the driver
Jun 12 14:54:58 <shubhangi> jusy probably*
Jun 12 14:55:18 <Abhishek_> against handling in userspace, do you have any link discussing that?
Jun 12 14:55:25 <shubhangi> yes
Jun 12 14:56:03 <shubhangi> sending it once i find it
Jun 12 14:56:06 <shubhangi> again
Jun 12 14:56:58 <Abhishek_> I see, also look into possible ways user-written modules can "plug into" pruss-remoteproc .
Jun 12 14:58:27 <shubhangi> okay..
Jun 12 15:03:20 <shubhangi> Abhishek_: http://yarchive.net/comp/linux/userspace_io.html
Jun 12 15:04:33 <shubhangi> Abhishek_: i read other arguments somewhere else too ... like if the user program is ended in the middle of handling, you do not have a chance to clear the interrput.. will sort of mess things up
Jun 12 15:05:08 <shubhangi> Abhishek_: a possible alternative is to allow users to run a callback function upon recieving an IRQ
Jun 12 15:05:40 <shubhangi> http://stackoverflow.com/questions/7986260/linux-interrupt-handling-in-user-space this too
Jun 12 15:05:45 <Abhishek_> yeah, so when I said interrupts, consider them "callbacks"
Jun 12 15:05:51 <shubhangi> okay
Jun 12 15:05:54 <Abhishek_> in userspace, that is
Jun 12 15:06:10 <shubhangi> development in progress (y)
Jun 12 15:06:17 <shubhangi> thanks
Jun 12 15:06:58 <shubhangi> Hows banglore by the way. I hear the weather's really good these days
Jun 12 15:08:23 <Abhishek_> The weather is good here, never really "felt the heat" , though there have been heat waves in most of the Northern part.
Jun 12 15:10:52 <shubhangi> Abhishek_: its unbearable up here
Jun 12 15:11:24 <Abhishek_> in Varanasi?
Jun 12 15:12:15 <shubhangi> at home right now. jabalpur
Jun 12 15:14:04 <Abhishek_> just hope the rains happen soon
Jun 12 15:14:23 <shubhangi> hope so
Jun 12 15:35:04 <shubhangi> Abhishek_: why does beaglelogic map kmalloc buffers to dma ? cant pru directly write to pa available from kmalloc
Jun 12 15:50:12 <apaar__> ds2 :ping
Jun 12 16:33:07 <nerdboy> moin (west coast)
Jun 12 16:35:24 <vvu> hello!
Jun 12 16:39:02 * nerdboy delays looking at neon assm bug
Jun 12 18:48:06 <Abhishek_> shubhangi: DMA mapping is necessary to ensure "cache coherency"
Jun 12 20:51:28 * nerdboy 's cache is definitely incoherent...
**** ENDING LOGGING AT Sat Jun 13 02:59:59 2015