Current Server Time: 17:04 (Central Europe)

#apertus IRC Channel Logs

2020/12/01

Timezone: UTC


23:21
anuejn
axiom-recorder can write cinema dng now :)
23:21
anuejn
and it can convert a raw8 / raw12 sequence to cinema dng
23:22
Bertl_oO
yay!
03:18
lexano
left the channel
03:23
Bertl_oO
off to bed now ... have a good one everyone!
03:23
Bertl_oO
changed nick to: Bertl_zZ
03:23
lexano
joined the channel
03:41
lexano
left the channel
05:40
BAndiT1983|away
changed nick to: BAndiT1983
07:10
se6ast1an
anuejn: very cool!
07:10
se6ast1an
from which sources so far?
07:27
vup
se6ast1an: currently it supports the tcp streaming we had for the micro, a folder of raw8 or raw12 files or a single raw8 / raw12 file
07:27
se6ast1an
great
09:48
anuejn
se6ast1an: could you check if the cinema dng the recorder produces can be opened by davinci?
09:48
anuejn
I can prepare you a folder somewhere
09:49
se6ast1an
sure
09:50
EmilJ
Hi all! I'm currently thinking if there's a better way to figure out DDR controller access efficiency than to assume that the "valid, not ready" or "ready, not valid" signals in a StreamInfo instance at the AXI_HP master port reflect it well
09:53
EmilJ
I went through the framebuffer experiment design and gathered this much: 1. the 16k async stream FIFO is kept full, so it's fine if we add latency to its influx of data by accumulating DDR read requests into bursts in the AxiReader
09:57
anuejn
EmilJ: yup added latency will bring nothing bad there :)
10:01
EmilJ
another way of profiling how much we are holding up DDR would be to just, give top priority to the PL accesses over the CPU, and then benchmark memory write and read worst case scenarios. But in that scenario, we're going to have "noise" from the kernel doing stuff as well, I bet
10:04
anuejn
se6ast1an: https://files.niemo.de/dng_test.zip
10:04
se6ast1an
got it
10:04
anuejn
:)
10:05
anuejn
EmilJ: that seems to be a reasonable approach
10:05
EmilJ
the CPU benchmark one?
10:05
anuejn
I would actually recommend to not go with the complete minimum but rather record some kind of histogram
10:06
anuejn
the one were you try to mimize cpu impact and then do throughput benchmarks
10:06
EmilJ
okay
10:07
EmilJ
I think this profiling isn't going to be of much use if I don't compare it with the reburster module
10:07
EmilJ
so maybe I should write that first
10:07
EmilJ
but it should be easy, since its output is a stream
10:08
se6ast1an
anuejn: dngs work fine in resolve
10:08
anuejn
nice
10:08
anuejn
is the framerate detected?
10:08
se6ast1an
the pixel errors on the right edge are unrelated I think
10:08
anuejn
(should be 24fps)
10:09
se6ast1an
indeed it detected 24FPS
10:09
anuejn
se6ast1an: yes the pixel errors are also present in the original demo footage :)
10:09
se6ast1an
4096x3072
10:09
anuejn
yaaay :)
10:09
anuejn
nice
10:10
se6ast1an
great work indeed!
10:10
anuejn
that is another step closer to a working recording setup :)
10:10
se6ast1an
hurray!
10:10
anuejn
sadly I have no way of using davinci anymore because they broke it for intel gpus on linux
10:11
anuejn
and my current computer doesnt have a dedicated gpu anymore
10:11
se6ast1an
I see
10:11
se6ast1an
photoshop also opens single dngs fine
10:11
anuejn
so if I want to do some real video editing I have to come up with a open source solution ;P
10:11
se6ast1an
says 8bits though for some reason...
10:12
anuejn
yeah I can open single dngs in raw therapee just fine
10:12
se6ast1an
might be related to output not input...
10:12
se6ast1an
great
10:12
anuejn
but that is... no fun
10:12
anuejn
hm... should be 12 bit
10:12
anuejn
at least the data in there is certainly 12bit
10:12
anuejn
but maybe resolve just uses 8?
10:13
se6ast1an
yeah, ingore photoshop
10:26
se6ast1an
cant tell with resolve, its just what photoshop said
10:26
se6ast1an
but as I said its probably output related, not input
10:38
lexano
joined the channel
10:59
BAndiT1983
was darktable and rawtherapee mentioned regarding the CDNG files?
10:59
BAndiT1983
worked in the past for comparison of own code
10:59
BAndiT1983
ah, found rawtherapee, was just written with whitespace
11:00
se6ast1an
just tried quoting sample metal laser cut parts with oshcut and laserboost
11:00
se6ast1an
oshcut is US shipping only unfortunately so far
11:03
BAndiT1983
which part is it for?
11:04
se6ast1an
but laserboost is in spain
11:05
se6ast1an
no particular part yet
11:05
se6ast1an
but for future applications
11:05
se6ast1an
but its great to see how cheap these have become
11:05
se6ast1an
parts are just a few $ each
11:06
BAndiT1983
nice!
11:07
se6ast1an
also bending included
11:10
BAndiT1983
how do you supply the data for bending?
11:12
se6ast1an
lines for the bend tool basically
11:12
se6ast1an
angle I need to check how to supply
11:12
BAndiT1983
that was the next question ;)
11:41
se6ast1an
maybe you define certain line styles as certain angles
11:41
se6ast1an
as all you supply is a dxf file
11:41
se6ast1an
like dashed line = 90°
11:49
BAndiT1983
have tried their example, but there is no option for bending at all for me
11:49
se6ast1an
laserboost or oshcut?
11:50
BAndiT1983
laserboost
11:50
se6ast1an
yeah I dont think they do bending
11:50
se6ast1an
oshcut does
11:50
BAndiT1983
yep, just opened the page, looks like the sheet metal planning in fusion 360
12:53
EmilJ
https://gitlab.com/tywonemi-school-stuff/zynq-hdmi-nmigen/-/blob/master/hdmi_framebuffer.pdf
12:53
EmilJ
you can see how at some point I gave up and just drew a black box when I couldn't be bothered to demonstrate what approximately it does
12:54
BAndiT1983
nice image!
12:54
EmilJ
also I have no idea how to display something like "this goes ready only if X" - I guess it's a bad idea and in docs should only be described in text because it obscures the actual data flow?
12:54
EmilJ
thanks
12:55
BAndiT1983
can't speak for the digital processing side, but i would keep diagrams according to their topic, so your case would need additional sequence diagram, at least we do it like that for software planning
12:56
Bertl_zZ
changed nick to: Bertl
12:56
Bertl
morning folks!
13:01
se6ast1an
good day
13:03
vup
EmilJ: nice diagram :)
14:46
aombk2
joined the channel
14:50
aombk
left the channel
14:52
RexOrCine
joined the channel
15:12
anuejn
yup indeed nice diagram :)
15:19
EmilJ
do you like the forbidden clock divider
15:20
EmilJ
I didn't have a better idea to visualize CLK/2 than to draw a flip flop with a feedback loop
15:38
anuejn
I see
15:38
anuejn
quite forbidden
15:38
anuejn
;)
16:44
Bertl
off for now ... bbl
16:45
Bertl
changed nick to: Bertl_oO
17:57
mumptai
joined the channel
18:17
Bertl_oO
EmilJ: what's the purpose of the FF-divider?
18:21
Bertl_oO
changed nick to: Bertl
18:22
EmilJ
well what the gearbox does is take one double word on its input and serve two words on its output sequentially
18:22
Bertl
I mean, why not use a BUFR there?
18:24
EmilJ
don't worry, there's not a literal flip flop there like that - it's just a simplified view of a bunch of conditional statements
18:24
EmilJ
which act as if the same as what I drew
18:26
Bertl
conditional statements on the clock line?
18:27
EmilJ
Nope, just on a counter's value, really
18:28
Bertl
well, interesting way to 'document' that then :)
18:28
EmilJ
yeah I should have just drawn a box with "/2" written on it instead
19:03
bert_bxl
joined the channel
19:03
bert_bxl69
joined the channel
19:08
bert_bxl
left the channel
19:08
bert_bxl69
left the channel
19:09
bert-bx
joined the channel
19:10
se6ast1an
Welcome bertand
19:11
bert-bx
hi Sebastian
19:12
Bertl
hey bert-bx! how is it going with the remote setup?
19:12
bert-bx
I think it's ready
19:12
Bertl
great! /msg me the details in private
19:25
RexOrCine
left the channel
19:25
RexOrCine
joined the channel
19:58
RexOrCine
left the channel
20:03
RexOrCine
joined the channel
21:04
aombk2
left the channel
21:09
aombk
joined the channel
21:42
Umori
left the channel
21:55
BAndiT1983
changed nick to: BAndiT1983|away
21:56
BAndiT1983|away
changed nick to: BAndiT1983
21:56
BAndiT1983
changed nick to: BAndiT1983|away
21:56
BAndiT1983|away
changed nick to: BAndiT1983
22:30
Umori
joined the channel
22:52
bert-bx
Thank you vup !
22:54
vup
you're welcome :)
22:56
BAndiT1983
changed nick to: BAndiT1983|away
22:57
bert-bx
thank you very much guys !
22:57
Bertl
you're welcome!
22:58
bert-bx
Next step - shooting !