Current Server Time: 01:44 (Central Europe)

#apertus IRC Channel Logs

2014/07/04

Timezone: UTC


01:19
fsteinel
left the channel
01:21
fsteinel
joined the channel
02:00
Bertl
off to bed now ... have a good one everyone!
06:07
jucar1
joined the channel
06:08
jucar
left the channel
07:46
jucar1
left the channel
07:55
se6astian
joined the channel
07:58
se6astian
good morning
08:02
jucar
joined the channel
10:37
Bertl
morning folks!
15:58
se6astian
left the channel
16:26
skinkie1
joined the channel
16:26
skinkie1
Bertl: would it be possible to debug the multiple programs fsbl problem this evening?
16:26
skinkie1
alternatively I'll setup a 32bit toolchain and see how far i can make it
16:31
intracube
joined the channel
17:39
Bertl
well, we can try, but note that I'm somewhat busy (doesn't mean I can't provide some input)
17:43
Bertl
it might be worth figuring out why your distro doesn allow LD_LIBRARY_PATH for the xilinx sdk
17:44
Bertl
one simple option though would be to create a namespace for the tools and put the libraries in the correct place (or even a chroot should do)
17:44
skinkie1
i'll see if i can just extract something like arch x86
17:44
skinkie1
and try the xilinx sdk under it
17:44
skinkie1
chroot exacty
17:46
Bertl
it should be simple if you can use the sdk to compare the output stages with objdump
17:46
Bertl
to figure out where the 'by hand' approach derails
17:46
skinkie1
ah you mean get something that works
17:47
skinkie1
ah :)
17:47
skinkie1
yep
17:47
skinkie1
i am really not deep in this level of 'detail'
17:47
skinkie1
especially debugging executable fileformats
17:48
skinkie1
i am more into reverse engineering
17:49
Bertl
not sure the axiom boot/debug system is the best project then :)
17:50
Bertl
you might have more fun with lens mount control and hdmi recorder compatibility - basically everywhere where we interface with the outside (i.e. proprietary interfaces)
17:51
Bertl
alternatively you can write a placer and router for the zynq :)
17:52
skinkie1
i think if we are passed the point of getting it to work
17:52
skinkie1
i am useful ;)
17:52
skinkie1
now the sdk is a blackbox to me
17:53
Bertl
well, as I said, I used it once on a late evening to create the FSBL, never since then (except for last time you asked me how it was created)
17:53
Bertl
the elphel folks are working on a FOSS FSBL btw
17:54
skinkie1
yeah but your comment included that you have also used the 'old' toolchain
17:54
skinkie1
yes, I might try that out too
17:54
Bertl
(I probably mentioned that already)
17:54
Bertl
but last time I checked it was rather incomplete
17:54
skinkie1
http://wiki.elphel.com/index.php?title=Ezynq
17:55
skinkie1
"Write level training for DDR memory doesn't work in MicroZed and ZC706" <- what does that mean?
17:57
Bertl
probably means that the DDR memory won't be useable
17:58
Bertl
IIRC, elphel doesn't plan on using the PS DDR interface on their camera
17:58
Bertl
i.e. they are creating their own DDR3 interface
17:58
skinkie1
sounds like Andrey found something us faster ;)
18:01
skinkie1
i'm but going for the chroot now
18:02
Bertl
us?
18:02
skinkie1
microsecond
18:03
Bertl
ah, I think the FSBL project started before we even started with the AXIOM Alpha
18:05
skinkie1
Andrey was working on the 373 years ago already
18:05
Bertl
if you feel like incorporating the elphel FSBL, go ahead, you might want/be-able to reverse engineer the ddr training from the xilinx fsbl and reimplement it properly on the elphel one
18:05
skinkie1
the 393 is more recent
18:07
skinkie1
the 373 was actually based on a Ti Davinci
18:07
skinkie1
but all the stuff for image processing was not used, that was the part andrey wanted to do himself
18:12
skinkie1
chroot works
18:12
skinkie1
Xilinx tools as well
18:40
skinkie1
Bertl: i have everything now running from a chroot
18:41
skinkie1
one of the first problem that i see from the Xilinx SDK is "fatal error: xil_io.h: No such file or..."
18:41
skinkie1
going to try if the hints of the forum works
18:41
skinkie1
but the pain is that nothing just seems to work out of the box
18:48
skinkie1
as magic... everything works
18:48
skinkie1
397098 Jul 4 17:47 apertusboot.elf
18:51
skinkie1
Bertl: I had one question you explained it before, but how is the devicetree defined?
19:05
skinkie1
http://forums.xilinx.com/t5/Embedded-Linux/device-tree-generate-error/td-p/482308 <- thats my current issue ;)
19:15
skinkie1
I have hacked the tcl script a bit, but i have no clue what slaveip should be doing ;)
19:15
skinkie1
but at least i get what i wanted
19:33
jucar
left the channel
20:03
Bertl
sorry got disconnected ... reading up
20:04
jucar
joined the channel
20:05
Bertl
okay, I have no clue regarding the device tree generation
20:05
Bertl
as I said, I created the devicetree myself (with vi)
20:05
skinkie1
how did you new about the contents ;)
20:05
skinkie1
anyway
20:05
skinkie1
i have to read how you did he u-boot-digilent config gaain
20:05
Bertl
I also have no clue what you hacked in 'the' tcl script
20:06
Bertl
I presume there is a commend? called slaveip? or so?
20:06
skinkie1
yeah something like it
20:06
Bertl
did you try help on it?
20:07
skinkie1
no not yet
20:07
skinkie1
it was somesort of default configuration
20:07
Bertl
next would be to check ug835 for documentation
20:07
skinkie1
but i am currently moving on with kernel/uboot etc
20:08
Bertl
yeah, I don't think that you will really need it
20:08
Bertl
probably some magic to incorporate some special config from the gui tools/wizzards
20:09
Bertl
the device tree is openly documented and the device tree is only used by the kernel
20:09
skinkie1
when i try your uboot version
20:09
Bertl
so it has to match for the kernel (syntax) and describe the hardware (semantic)
20:09
skinkie1
i get a lot undefined reference stuff
20:09
Bertl
which uboot version are you using?
20:09
skinkie1
the u-boot-digilent
20:10
Bertl
branch/version/commit?
20:10
skinkie1
b55d4b1bbd5e6a98838414949fae620d34098f0a
20:10
skinkie1
basically the same as last time
20:11
skinkie1
but now compiled with a different gcc/linker
20:11
Bertl
okay, example for the undefined reference stuff?
20:13
skinkie1
for your info the xilinx version works out of the box with compilation
20:13
Bertl
okay, and you are using the xilinx compiler or your own?
20:14
skinkie1
everything xilinx
20:14
skinkie1
because i want to make progress...
20:14
Bertl
okay, hehe :) I used my own cross compiler toolchain for u-boot
20:15
skinkie1
http://pastebin.com/bRccy5Uj
20:15
Bertl
i.e. the default arm-gcc available for my distro
20:15
skinkie1
notice the .rel.plt stuff
20:15
skinkie1
we saw that incompatibility before
20:18
Bertl
yeah, I suspect that is the slightly too new/too eager to optimize compiler version
20:19
Bertl
and you said, the xilinx u-boot version compiles just fine?
20:29
skinkie1
yes
20:30
Bertl
then I'm pretty sure that xilinx patched the relevant parts in their tree
20:30
Bertl
I would look into a diff between the two u-boot branches and watch out for makefile changes
20:31
Bertl
IMHO it is only a matter of reordering and/or adding certain commands/files
20:34
Bertl
but you can test with the xilinx version as well, it shouldn't affect you much as I suspect you won't be using TFTP right now
20:34
Bertl
(and maybe they even fixed the performance issue there since I tested)
20:35
Bertl
also note that you want to use a microzed specific config if there is one by now
20:35
skinkie1
ok, so i have now probably everything required
20:35
skinkie1
uImage, uBoot, the fsbl, devicetree
20:36
Bertl
you can take one of the kernels I built for a start
20:36
skinkie1
i leave the ram disk for now
20:36
skinkie1
i took the stock configuration for zynq
20:36
Bertl
I will update the kernel build soon because I have to test out a few things
20:36
Bertl
yeah, you can use that as well for testing
20:37
Bertl
(it might not work with my devicetree though :)
20:43
skinkie1
I have a boot.bin :)
20:43
skinkie1
obviously clueless if it works
20:43
Bertl
go ahead, test it
20:44
skinkie1
after some relaxing i'll test it... very hot :)
00:23
jucar
left the channel
00:25
jucar
joined the channel
00:55
jucar
left the channel
00:57
jucar
joined the channel