Current Server Time: 13:35 (Central Europe)

#apertus IRC Channel Logs

2013/09/07

Timezone: UTC


00:39
aombk
joined the channel
00:39
aombk
left the channel
00:39
aombk
joined the channel
00:40
aombk
left the channel
00:40
aombk
joined the channel
00:40
aombk
left the channel
00:40
aombk
joined the channel
04:24
aombk
left the channel
05:45
sasha-w
joined the channel
05:45
sasha-w
left the channel
09:01
sasha-w
joined the channel
09:01
sasha-w
left the channel
12:25
Bertl
morning everyone!
13:19
sasha-w
joined the channel
13:19
sasha-w
left the channel
17:49
se6astian
joined the channel
17:49
se6astian
evening
20:35
se6astian
left the channel
21:01
gabe
joined the channel
21:03
gabe
left the channel
21:07
gcolburn
joined the channel
21:07
gcolburn
hello
21:08
Bertl
hello gcolburn!
21:08
gcolburn
This is gabe, you helped me with the zed board about a week ago
21:08
gcolburn
finally got an IRC client
21:09
Bertl
yup, figured that (the identity) from the previous connect. congrats for getting an IRC client!
21:09
gcolburn
so I've been going through the speedway tutorials and have been able to write custom HDL and write bare metals apps to connect to peripherals, so I'm getting more familiar with the architecture
21:09
Bertl
excellent!
21:09
gcolburn
there's an aspect not completely clear in the tutorials that I was hoping you can clarify.
21:10
Bertl
let's hear
21:10
gcolburn
it looks to me that in order to communicate between the ARM and the PL, your options are basically 2 master AXI ports, two slaves, and the several high performance AXI ports
21:11
gcolburn
with the master/slave AXI ports having up to 32 bits each
21:11
Bertl
basically, but there are other overlaps and channels as well
21:12
gcolburn
what do you plan on using for interfacing with the sensor?
21:12
Bertl
you mean, getting data from the sensor into the PS, yes?
21:12
gcolburn
yes
21:13
gcolburn
the high performance AXI bus? (64 bit)?
21:13
Bertl
I think the best option there is DMA to the DDR Memory
21:13
gcolburn
ok. I haven't played with that yet
21:14
Bertl
the main processing in the prototype will happen int the PL
21:14
Bertl
so the PS is only involved when we want to record or capture still images
21:15
Bertl
i.e. the normal data flow will be sensor -> PL -> hdmi
21:15
gcolburn
ok, just by reading out each frame from DDR?
21:15
gcolburn
(for the still images)
21:16
Bertl
for example, as the ARM cores have direct access to the DDR, it shouldn't be too hard to get a driver for that
21:16
gcolburn
yeah
21:17
Bertl
but I haven't spent much thought on the capture part yet, as I consider it not too problematic
21:18
gcolburn
I know this is long term (after the prototype), but have you guys talked about options for implementing the user interfaces? would you go with something like android or a more simpler linux and write more custom UI code?
21:19
Bertl
you are right, this is long term atm :)
21:19
Bertl
I can imagine running two ARM cores with two different systems
21:19
Bertl
for example, one android and the other bare metal (or normal linux)
21:20
Bertl
where the android one uses a 'framebuffer' like interface for an overlay
21:20
Bertl
or we can do the actual (G)UI as bare metal and have a linux running for streaming and coordination
21:21
gcolburn
interesting
21:22
gcolburn
i don't know if sebastian told you much about what my interest in this project is.
21:22
gcolburn
its very similar to a project I've wanted to do for a number of years
21:22
Bertl
well, won't hurt to repeat it here for others as well
21:23
gcolburn
and by learning and helping out with this project it'll enable me to do my own project eventually
21:24
gcolburn
one of my hobbies is landscape photography, both digital and large format (with film)
21:24
gcolburn
so I've wanted to create a digital back form my large format camera for much cheaper than digital medium format backs
21:24
gcolburn
and have full control over the sensor
21:25
Bertl
cmosis has a nice sensor for this kind of photography I think
21:25
gcolburn
oh yeah?
21:25
Bertl
http://www.cmosis.com/index.php?/products/standard_products/chr70m
21:27
gcolburn
that's cool. I didn't know they had a sensor that large
21:27
gcolburn
the issue I had before is there are few manufacturers out there that make APS-C and larger sensors that an individual can go out and purchase
21:27
Bertl
it needs AD converters though, so no direct digital
21:28
gcolburn
there's been one that Aptina as advertised for a few years but wasn't available for me to purchase. when I found out about apertus I saw the CMOSIS sensor for the project that I had never seen before
21:29
gcolburn
yeah, I could eventually switch to that sensor after I get a prototype working
21:30
gcolburn
has CMOSIS been pretty good to work with for getting quotes and datasheets?
21:30
Bertl
maybe you can even use the prototype for your purpose
21:31
Bertl
you won't need high data rates for landscape, no?
21:31
gcolburn
yeah, that's what I was thinking (to use the prototype) and hook up an LCD display and work on the user interface
21:31
gcolburn
not particularly
21:31
gcolburn
high data rates that is
21:31
Bertl
yeah, and the planned setup should not restrict the sensor in anything but the high frame/data rates
21:32
Bertl
so every trick you can do with the sensor configuration can be done
21:32
gcolburn
cool
21:32
gcolburn
the sensor has interesting features like the dual-exposures and adjustable response curves (with the 3 linear regions)
21:33
Bertl
yes, and I'm very happy/excited about the piplined global shutter
21:38
gcolburn
by the way, last week when I was booting to linux with a custom PL bit stream (for example the blinking LEDs), I wasn't able to get to the console prompt, the PL would download and run correctly. I assume this expected?
21:39
Bertl
hmm, no, the PS should work fine without any PL code
21:39
gcolburn
really? hmm
21:39
Bertl
actually the PS starts prior to loading PL code
21:39
gcolburn
because I'd switch the PL out to the original one and it would work fine
21:39
Bertl
(mainly for security reasons)
21:39
Bertl
but there could be lockups when the PS code requires some PL interfaces
21:40
gcolburn
that's what I was thinking
21:40
Bertl
for example, visualize a setup where the PL contains an AMBA interface used by a driver loaded in the PS
21:40
Bertl
then you switch the PL and *poof* the hardware is gone
21:41
Bertl
but OTOH, a properly written driver should handle that gracefully
21:41
Bertl
but I actually think you are observing something completely different here :)
21:42
Bertl
when you connect to the jtag, many tools stop the PS
21:42
Bertl
from the console, it looks like the system has frozen
21:42
gcolburn
the output was actually the same as when I was getting the "file not found" error which turned about to be a permission issue on the symbolically linked bin file. so the last few lines are:
21:42
gcolburn
[ 0.430000] ###############################################
21:42
Bertl
continuing execution will make it work again
21:42
gcolburn
[ 0.430000] # #
21:42
gcolburn
[ 0.440000] # Board ZED Init #
21:42
gcolburn
[ 0.440000] # #
21:42
gcolburn
[ 0.440000] ###############################################
21:42
gcolburn
[ 0.440000]
21:42
gcolburn
[ 0.450000] hw-breakpoint: found 5 (+1 reserved) breakpoint and 1 watchpoint.
21:43
gcolburn
[ 0.450000] hw-breakpoint: maximum watchpoint size is 4 bytes.
21:43
gcolburn
[ 0.480000] xslcr xslcr.0: at 0xF8000000 mapped to 0xE0808000
21:43
gcolburn
[ 0.490000] bio: create slab <bio-0> at 0
21:43
gcolburn
[ 0.490000] gpiochip_add: registered GPIOs 0 to 245 on device: xgpiops
21:43
gcolburn
[ 0.490000] xgpiops e000a000.gpio: gpio at 0xe000a000 mapped to 0xe080a000
21:43
Bertl
(pastebin :)
21:43
Bertl
so, did you successfully boot into a linux system yet?
21:44
gcolburn
I can
21:44
gcolburn
just not with a custom PL
21:44
gcolburn
like the blinking LEDs
21:44
Bertl
okay
21:44
Bertl
what kernel are you using?
21:44
gcolburn
the one from the apertus repository
21:45
Bertl
from the TFTP dir?
21:45
Bertl
the ZedBoard_OOB_Design?
21:45
gcolburn
yeah
21:46
Bertl
okay, that will be picky, let me upload a better one for this purpose
21:47
gcolburn
ok, that would be good
21:48
gcolburn
I was working on finishing the speedway tutorials with the bare metal apps, and then switch to the tutorials on boot loaders, building linux kernels, drivers, apps, etc. I watched the videos, just need to do the labs
21:55
Bertl
upload will take a little, but there is a BERTL.PL and BERTL.xilinx here http://vserver.13thfloor.at/Stuff/AXIOM/TFTP/ZED/
21:56
Bertl
the BERTL.PL contains the devicetree I currently use (which boots into the flash)
21:56
Bertl
and the BERTL.xilinx will soon contain the rebuilt zImage based on the xilinx git tree
21:56
Bertl
(which has all the relevant features enabled but doesn't expect any PL code)
21:58
Bertl
my symbolic links are those:
21:58
Bertl
devicetree_sdp2.dtb -> BERTL.PL/devicetree.dtb
21:58
Bertl
zImage -> BERTL.xilinx/zImage
21:59
Bertl
note that the sdp2 means second partition on the flash card, but you can adapt that for the ram filesystem as well