| 23:00 | Rebelj12a | I mean it is gpu based but to run on a phone gpu it would have to be pretty light I would think
|
| 23:00 | wescotte | Phone GPUs are pretty powerful
|
| 23:01 | Rebelj12a | Hm
|
| 23:01 | wescotte | Look at the games you can play on them
|
| 23:02 | Rebelj12a | true true
|
| 23:02 | wescotte | I bet the typical phone GPU is more powerful than the PS2/Original XBox GPUs
|
| 23:03 | wescotte | Bertl: Well, now I have an excuse to figure out how to use the lab...
|
| 23:03 | Rebelj12a | Oh yeah most likely, doesnt mean they can interpret the shaders and physics as well but thats cpu processing although phones are getting really good at handing that off to gpu
|
| 23:03 | Rebelj12a | Haha
|
| 23:05 | Rebelj12a | Bah yeah nevermind, research yielded opengl as the best
|
| 23:06 | Rebelj12a | Is it a pic32mx or MZ EC? or something else entirely?
|
| 23:07 | Rebelj12a | bah ill check the wiki
|
| 23:08 | Bertl | PIC32MZ
|
| 23:08 | Rebelj12a | mz ok
|
| 23:08 | Rebelj12a | On Arch Linux good choice
|
| 23:08 | Bertl | probably on bare metal
|
| 23:09 | Bertl | Arch will be running on the FPGA core(s) though
|
| 23:23 | wescotte | How long does it typically take to program an FPGA? Is it a static amount of time or does it depend on the complexity of the code?
|
| 23:24 | Bertl | what exactly do you mean with 'program'?
|
| 23:25 | wescotte | Is it like blitting to a frame buffer but instead of pixels it's enabling/disabling gates?
|
| 23:26 | Bertl | somewhat similar, the gates are all there and in fixed places, also the 'wiring'. what you configure is basically connections on intersecting wires
|
| 23:27 | Bertl | and lookup tables
|
| 23:27 | wescotte | I don't know much about them.. Do you change the lookup tables on the fly too?
|
| 23:28 | wescotte | Or does that get too complicated?
|
| 23:28 | Bertl | they can be changes by a mechanism called dynamic reconfiguration, which is supported on some FPGAs
|
| 23:28 | Bertl | *changed
|
| 23:29 | Bertl | basically all but the simplest gates are actually lookup tables
|
| 23:29 | wescotte | So you can basically do almost an inturrupt and reprogram the entire thing to do a specific task and once done revert back the original design?
|
| 23:30 | Bertl | if you don't mind that in the meantime everything else won't work, yes
|
| 23:30 | wescotte | So you could build complex instructions as you need them
|
| 23:31 | wescotte | I just wasn't sure how fast you could reprogram it completely
|
| 23:31 | Bertl | the FPGA is not a processor, it is more like a breadboard
|
| 23:31 | wescotte | right
|
| 23:32 | Bertl | but you can 'build' a processor in the FPGA
|
| 23:32 | Bertl | which then can execute complex instructions
|
| 23:33 | wescotte | I just was curious how you can "wire the breadboard"
|
| 23:33 | Bertl | by uploading a bitstream :)
|
| 23:34 | Rebelj12a | Shouldnt.... shouldnt have joined the lab.... tooo many things to post..........
|
| 23:35 | wescotte | Bertl: okay, wasn't sure if there was some physical "reaction" that limited how fast it could go. Sorta like a CD-rW erasing/writing where it needs a specific amount of light for a specific duration before it can flip
|
| 23:36 | Bertl | well, yes, it takes some time as the bandwidth for uploading the bitstream is not that great
|
| 23:36 | wescotte | Bertl: So it's basically a grid of gates that are all physically wired together and via a lookup table determines the actual paths?
|
| 23:37 | Bertl | basically there are a huge number of mosfet transistors, which are controlled by a bitstream
|
| 23:37 | Bertl | all of them create connections in the existing fabric
|
| 23:38 | Bertl | parts of the fabric are 'wires', other parts are lookup tables (they are forming the gates) and again other parts are hardened parts (i.e. special electronics)
|
| 23:40 | Bertl | http://www3.ntu.edu.sg/home2010/vipin2/publications/esl2013-vipin.pdf
|
| 23:40 | Bertl | (some info regarding efficient partial reconfiguration)
|
| 23:43 | wescotte | I guess what I was originally trying to ask was if it's common for FGPAs to create new instructions on the fly.. A compiler will optimize chunks of code to use SIMDs if they exist. So if you have code to detect that sort of thing for your FGPA it could create that instruction on the fly and use it
|
| 23:44 | wescotte | Or do you have to writ the whole thing at once
|
| 23:44 | wescotte | I apologize if I'm not making any sense :)
|
| 23:45 | Bertl | I know what you mean but it doesn't make much sense :)
|
| 23:45 | Bertl | think of the FPGA like this:
|
| 23:45 | Bertl | you have on one side 64 inputs from the sensor
|
| 23:46 | Bertl | and on the other side 4 outputs to an HDMI connector
|
| 23:46 | wescotte | I think my mistake is trying to implement a cpu on a FGPA. They're more about single task then general purpose?
|
| 23:47 | Bertl | now you design a circuit (not a program) to retrieve the data from the sensor, reorder it, encode it and transmit it with a different data rate over those four outpus
|
| 23:47 | Bertl | *outputs
|
| 23:47 | Bertl | there is no program, unless you create a soft cpu (as circuit) in the FPGA
|
| 23:48 | Bertl | in theory, that soft cpu could use the reconfiguration interface to do some kind of just in time SIMD reconfiguration to 'speed up' processing
|
| 23:49 | Bertl | practically the overhead for this would be way larger than doing the same with a normal CPU or even better with predefined circuits in the FPGA
|
| 23:50 | wescotte | So above when you said "Arch will be running on the FPGA core(s)" you are creating a soft cpu for that?
|
| 23:50 | Bertl | no, there are two hardened arm cores in the zynq
|
| 23:51 | Bertl | we will be using those, but we could create a soft CPU as well
|
| 23:51 | wescotte | okay I see now
|
| 23:51 | Bertl | (only it would waste FPGA resources in our case)
|
| 23:52 | wescotte | So what tasks are you intending to use the FPGAs for? Taking the sensor data, debayering, and creating the 3 HDMI streams?
|
| 23:54 | Bertl | for example
|
| 00:00 | wescotte | I have a lot of catching up to do. I've been programming as a hobby for at least 25 years now but I keep getting further and further away from the metal.
|
| 00:02 | wescotte | which is why i was originally asking about UI stuff. I might actually be able to contribute there :)
|
| 00:08 | Bertl | well, if you can get comfortable with designing UI stuff in C, then why not
|
| 00:17 | wescotte | Cuz that's no fun :)
|
| 00:17 | Rebelj12a | haha
|
| 00:29 | FergusL | left the channel |
| 00:30 | Bertl | Elite had a quite spectacular GUI and '3D' rendering with way less than we have on the hardware side :)
|
| 00:32 | Rebelj12a2 | joined the channel |
| 00:34 | Rebelj12a | left the channel |
| 00:34 | Rebelj12a2 | hmm
|
| 00:38 | intracube | left the channel |
| 00:55 | Rebelj12a2 | left the channel |
| 00:55 | Rebelj12a | joined the channel |
| 01:03 | Rebelj12a2 | joined the channel |
| 01:04 | Rebelj12a | left the channel |
| 01:13 | Rebelj12a2 | left the channel |
| 01:13 | Rebelj12a | joined the channel |
| 01:56 | FergusL | joined the channel |
| 01:59 | intracube | joined the channel |
| 02:12 | __anton__ | left the channel |
| 02:22 | wescotte | left the channel |
| 03:11 | Rebelj12a | left the channel |
| 03:11 | Rebelj12a | joined the channel |
| 03:12 | Bertl | off to bed now ... have a good one everyone!
|
| 03:12 | Bertl | changed nick to: Bertl_zZ
|
| 04:24 | Rebelj12a | left the channel |
| 04:25 | Rebelj12a | joined the channel |
| 04:59 | Colinelves | joined the channel |
| 04:59 | Colinelves | left the channel |
| 05:06 | Colinelves | joined the channel |
| 05:13 | Colinelves | left the channel |
| 05:24 | Rebelj12a | left the channel |
| 05:25 | Rebelj12a | joined the channel |
| 06:09 | sb0_ | joined the channel |
| 07:13 | Rebelj12a | left the channel |
| 07:13 | Rebelj12a | joined the channel |
| 07:28 | danieel | left the channel |
| 07:38 | sb0_ | left the channel |
| 07:39 | danieel | joined the channel |
| 07:45 | se6astian|away | changed nick to: se6astian
|
| 07:46 | se6astian | good morning
|
| 08:02 | danieel | hi seb.. what does parking in front of hofburg cost? (kurzparkzone)
|
| 08:05 | se6astian | lol
|
| 08:06 | danieel | dont lol, there is a place to park there
|
| 08:07 | se6astian | 2€ per hour, max parking duration 2 hours
|
| 08:07 | se6astian | you need to buy the colored parkcertificates first though
|
| 08:07 | danieel | thanks, so i assume I will need a garage then
|
| 08:08 | se6astian | will cost you a multitude though
|
| 08:08 | danieel | staying like 2 days.. any better idea where to leave the car?
|
| 08:08 | danieel | yes i found some in 40e/day range
|
| 08:08 | se6astian | do you need the car during that time?
|
| 08:08 | danieel | no, i just unpack the stuff there and the car can go anywhere
|
| 08:09 | danieel | unpaid zone wont be near right?
|
| 08:09 | se6astian | then bring it to an outer district where parking is free
|
| 08:09 | danieel | best tip, with ubahn connection there?
|
| 08:10 | se6astian | let me check
|
| 08:10 | danieel | do you have those P+R sites there?
|
| 08:10 | danieel | paid with a ticket valid for mass transportation
|
| 08:10 | se6astian | yes, but they are not free most of the time either
|
| 08:10 | danieel | well, still will need the ticket to get back to central
|
| 08:10 | se6astian | they are typically located at the outskirts of vienna so people can communte into the city
|
| 08:12 | se6astian | do you have a GPS in your car?
|
| 08:13 | danieel | yes
|
| 08:13 | danieel | would be lost without it :)
|
| 08:13 | se6astian | bring your car to "Boschstraße 24"
|
| 08:13 | se6astian | its at the end of metro line U4 just around the corner
|
| 08:13 | se6astian | in the 18th district which has free parking
|
| 08:15 | se6astian | you will have to search a bit for a free space though
|
| 08:16 | davidak | joined the channel |
| 08:16 | danieel | thanks
|
| 08:17 | danieel | the U lines are more like train or a subway?
|
| 08:18 | se6astian | subway
|
| 08:18 | se6astian | the U stands for underground
|
| 08:18 | se6astian | even though some lines dont go underground all the way
|
| 08:27 | Rebelj12a | left the channel |
| 09:47 | intracube | left the channel |
| 10:06 | Bertl_zZ | changed nick to: Bertl
|
| 10:06 | Bertl | morning folks!
|
| 10:18 | se6astian | hey there
|
| 10:18 | se6astian | I think I am in love with the tormach mill :D
|
| 10:18 | se6astian | they even have an accessory for small volume injection molding
|
| 10:18 | se6astian | for just 800$
|
| 10:18 | se6astian | http://www.tormach.com/store/index.php?app=ecom&ns=prodshow&ref=32079
|
| 10:19 | se6astian | only downside I found so far is that their automatic toolchanger system is not compatible with EMC2
|
| 10:19 | se6astian | they clearly state that on the tool changer page
|
| 10:29 | se6astian | gotta go
|
| 10:29 | se6astian | changed nick to: se6astian|away
|
| 11:57 | aombk | "needs triage" in phabrikator is a high priority?
|
| 12:19 | Bertl | not sure, but it basically means that somebody should look at it and classify/assign
|
| 12:19 | Bertl | (at least to me :)
|
| 12:20 | Bertl | we might know more after tomorrow
|
| 13:00 | FergusL | left the channel |
| 13:36 | surami | joined the channel |
| 14:15 | surami | left the channel |
| 14:37 | Rebelj12a | joined the channel |
| 15:15 | Rebelj12a | Hey was it on anyone's mind, or discussed, compression of dngs and format?
|
| 15:15 | Rebelj12a | Well at least for archival.
|
| 15:19 | Rebelj12a | I mean I'm sure it will come since ML will develop for axiom, however a post from the MLRawviewer creator on archival dng http://thndl.com/how-dng-compresses-raw-data-with-lossless-jpeg92.html
|
| 15:20 | intracube | joined the channel |
| 15:22 | surami | joined the channel |
| 15:41 | surami | left the channel |
| 15:51 | legendin | joined the channel |
| 16:04 | Rebelj12a2 | joined the channel |
| 16:06 | Rebelj12a | left the channel |
| 16:08 | Rebelj12a2 | left the channel |
| 16:15 | Rebelj12a | joined the channel |
| 16:23 | aombk | Bertl, so there is a maximum number of shields that can be added to the board the beta will use?
|
| 16:24 | Rebelj12a | Well.. for size constraints at the very least functionality id say so haha
|
| 16:24 | Rebelj12a | Giant 4 ft long camera XD
|
| 16:25 | Bertl | aombk: in the current design, yes, precisely two shields, one on the left and one on the right side
|
| 16:26 | aombk | because of the enclosure size or because of board i/o connections?
|
| 16:27 | Bertl | because of the board I/O
|
| 16:27 | Bertl | in theory you could stack some of them up
|
| 16:28 | Bertl | i.e. make connectors on both sides of the shield, pass through some lanes and stack more than one shield on each side
|
| 16:28 | Rebelj12a | Power constraints and loss due to more shields?
|
| 16:29 | Bertl | yeah, all within reason, of course
|
| 16:29 | Rebelj12a | of course
|
| 16:29 | Bertl | and as I said, the 'lower' shields would have to be special for that to work
|
| 16:33 | aombk | and what shields are you planning to make?
|
| 16:34 | Bertl | at least one tripe low speed hdmi shield, and one high speed hdmi one
|
| 16:34 | Bertl | a debug shield and probably some button, key, LCD shield
|
| 16:35 | Rebelj12a | Oh debug shield I likes
|
| 16:35 | aombk | usb,wifi,sdcard are onboard?
|
| 16:38 | aombk | i mean no extra shield?
|
| 16:39 | aombk | also Accelerometer, magnetometer and gyroscope?
|
| 16:44 | aombk | bertl, you say "Highly customizable via modular I/O addons" that means after you make the basic i/o shields you have planned (and probably after release), you are going to start discussion on other shield ideas?
|
| 16:54 | Bertl | usb and ethernet, as well as internal SD card are onboard the MicroZed
|
| 16:54 | danieel | Bertl: what is the approximate volume of the shield can consume? ( X x Y x Z )
|
| 16:54 | Bertl | wifi can be done via usb, IMU is onboard the Beta board
|
| 16:57 | Bertl | the size over the Beta board is roughly 22 x 57mm, the height is at least 10mm probably more like 20mm
|
| 16:57 | Bertl | but you could extend the shield to the side if the case allows that
|
| 17:14 | philippej | joined the channel |
| 17:18 | pizthewiz | joined the channel |
| 17:25 | se6astian|away | changed nick to: se6astian
|
| 17:32 | se6astian | good evening
|
| 17:32 | se6astian | discussing T3
|
| 17:34 | philippej | bot will be back when some features are ironed out
|
| 17:34 | se6astian | ok :)
|
| 17:38 | Bertl | how are T3 and the bot related? :)
|
| 17:52 | Rebelj12a2 | joined the channel |
| 17:52 | Rebelj12a2 | left the channel |
| 17:58 | slikdigit | joined the channel |
| 17:58 | slikdigit | left the channel |
| 17:58 | slikdigit | joined the channel |
| 18:16 | philippej | left the channel |
| 18:23 | g3gg0 | joined the channel |
| 18:39 | legendin | left the channel |
| 18:43 | mithro | left the channel |
| 18:44 | mithro | joined the channel |
| 18:45 | FergusL | joined the channel |
| 18:52 | aombk2 | joined the channel |
| 18:53 | aombk | left the channel |
| 19:01 | philippej | joined the channel |
| 19:04 | aombk2 | left the channel |
| 19:05 | aombk2 | joined the channel |
| 19:32 | aombk3 | joined the channel |
| 19:32 | aombk2 | left the channel |
| 19:37 | se6astian | changed nick to: se6astian|away
|
| 19:45 | Bertl | off to bed now ... have a good one everyone!
|
| 19:46 | philippej | good night!
|
| 19:46 | Bertl | changed nick to: Bertl_zZ
|
| 20:08 | wescotte | joined the channel |
| 20:09 | Rebelj12a2 | joined the channel |
| 20:15 | danieel | left the channel |
| 20:25 | philippej_ | joined the channel |
| 20:29 | philippej | left the channel |
| 20:30 | philippej_ | left the channel |
| 20:36 | se6astian|away | changed nick to: se6astian
|
| 20:42 | Rebelj12a2 | left the channel |
| 20:58 | FergusL | left the channel |
| 21:00 | danieel | joined the channel |
| 22:48 | FergusL | joined the channel |
| 22:52 | se6astian | changed nick to: se6astian|away
|