| lemay | are there any plans for the milkymist involving GPS? | 01:58 |
|---|---|---|
| wolfspraul | right now we just hookup the sige front-end rf ic and see whether we can get it to work | 01:58 |
| wolfspraul | no big strategy or product plans | 01:58 |
| wolfspraul | sige sent us one evb, which went to kristianpaul who hooked it up to his milkymist one | 01:59 |
| wolfspraul | lemay: have you worked with namuru or osgps before? | 02:01 |
| lemay | No - never heard f either | 02:06 |
| lemay | I've done a lot of work with GPS though | 02:07 |
| lemay | from the RF front end to the tracking loops | 02:07 |
| wolfspraul | great | 02:07 |
| wolfspraul | if you stay here a little and timezone wise you and kristianpaul are both up I'm sure he might have a question for you | 02:08 |
| wolfspraul | seems he's not online right now, don't know what time it is in Colombia... | 02:09 |
| lemay | cool :) | 02:09 |
| lemay | I'd be happy to chat | 02:09 |
| qi-bot | [commit] Xiangfu Liu: Nanonote: ignore HUP signal (master) http://qi-hw.com/p/gmenu2x/64108ab | 02:10 |
| qwebirc86657 | Review of KiCAD http://dangerousprototypes.com/2011/08/16/kicad-open-source-pcb-layout-package-reviewed/ | 02:38 |
| qi-bot | [commit] Xiangfu Liu: gmenu2x, update and immediately stop then poweroff (master) http://qi-hw.com/p/openwrt-packages/d320abc | 02:40 |
| wolfspraul | qwebirc86657: thanks for sharing! | 02:48 |
| qi-bot | The build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08152011-2339/ | 02:52 |
| kristianpaul | hello lemay ! :) | 02:59 |
| kristianpaul | wich kind of tracking loops had you worked with? | 03:00 |
| wolfspraul | kristianpaul: he's out to dinner now :-) | 03:01 |
| kristianpaul | oh, ok | 03:01 |
| kristianpaul | quick update, them | 03:08 |
| kristianpaul | lemay: namuru correlator (actually kinda same tp gp2021 from zarlink) is ported now to milkymist | 03:08 |
| kristianpaul | missing and not implemented yet in software a proper tracking and pull-in algorythm that finally allow us to get navigation data :) | 03:09 |
| kristianpaul | later care about fixes :) | 03:09 |
| kristianpaul | so osgps wich stands to open source gps, was developd/used ten years or so ago with a gp2021 chip hookedup to a ISA card on a 100mhz computer | 03:12 |
| kristianpaul | the code is not the best you see, i'm not a skilled coder, but still dificult to get understan how it works as too many things are in the same file.. well | 03:12 |
| kristianpaul | osgps seems to implement a PLL-like methog for tracking | 03:13 |
| kristianpaul | dunno how it works in detail | 03:13 |
| kristianpaul | and btw how yo get to here ;)? | 03:14 |
| lemay | hah, I have to dust off those memories | 03:14 |
| lemay | I bumped into wolfspraul | 03:14 |
| lemay | :) | 03:14 |
| lemay | well, the tracking loop I've done was pretty much like a PLL as well | 03:15 |
| kristianpaul | http://www.colorado.edu/geography/gcraft/notes/gps/gif/bitsanim.gif | 03:15 |
| lemay | yea, correaltion | 03:16 |
| lemay | neat gif | 03:16 |
| kristianpaul | that above acording to the gp2021 datasheet can easilly achieved by a loop and namuru i guess | 03:16 |
| kristianpaul | i'll try that this weekend or earlier | 03:16 |
| kristianpaul | but | 03:16 |
| kristianpaul | at least graph it with gnuplot or similar | 03:17 |
| kristianpaul | once i had a clear correlatio peak | 03:17 |
| kristianpaul | the hunting star in order to pull-in http://home.earthlink.net/~cwkelley/receiv3.gif | 03:18 |
| kristianpaul | s/star/start | 03:18 |
| kristianpaul | and i'm not compley aware or understand a proper algortym for it. | 03:19 |
| kristianpaul | of course i'm now there ir no silver bullet on that as it depents on the aplication | 03:20 |
| lemay | Does the namuru have a complete RF front end? | 03:20 |
| kristianpaul | nope just correlator | 03:20 |
| kristianpaul | actually it still need aid from software to work, is not a complete base band solution | 03:20 |
| kristianpaul | here is where milkymist soc plays a role :) | 03:21 |
| lemay | what are those sma's for? IF? | 03:21 |
| kristianpaul | current frontend is a SE4162 from SiGE | 03:21 |
| kristianpaul | ir provides real and complex samples | 03:21 |
| kristianpaul | currently working on real mode, 16Mhz sampling clokc 2.556 IF and 2 bit Sign/mag out | 03:22 |
| kristianpaul | lemay: wich sma's exacly? | 03:22 |
| kristianpaul | from wich pic? | 03:22 |
| lemay | heh, right - http://www.dynamics.co.nz/index.php?main_page=page&id=9 | 03:23 |
| lemay | I forgot I got there from google and not from you showing me that | 03:23 |
| kristianpaul | oh wait, le me drop some links | 03:23 |
| kristianpaul | www.dynamics.co.nz/media/Namuru_GPS_datasheet.pdf <- namuru datasheet | 03:24 |
| kristianpaul | this is the EVB that wolfspraul talked about http://en.qi-hardware.com/wiki/File:SiGE_EVB_Zoomed.jpg | 03:24 |
| kristianpaul | here some sumary http://en.qi-hardware.com/wiki/GPS_Free_Stack | 03:25 |
| kristianpaul | lemay: was your work related to full software processing? | 03:25 |
| kristianpaul | was it offline or "real time" processing? | 03:26 |
| lemay | real time | 03:26 |
| kristianpaul | yay ! | 03:26 |
| kristianpaul | so far the _only_ achievement about signaling was confirm code in the signal i sampled for offline prossing | 03:27 |
| kristianpaul | afaik i never got track, i guess because i wasnt aware osgps dint worked with complex samples.. | 03:28 |
| lemay | you successfully correlated CA code in post-processing? | 03:28 |
| kristianpaul | with namuru? | 03:28 |
| lemay | uh, heh | 03:29 |
| kristianpaul | in short words i still on acquisition | 03:29 |
| lemay | right | 03:29 |
| lemay | what sort of data are you working with? | 03:30 |
| kristianpaul | from namuru? | 03:31 |
| lemay | I mean the RF | 03:32 |
| kristianpaul | currently sige rf fronted provides real samples, 2 bit sig/mag | 03:32 |
| kristianpaul | 2.556Mhz IF 16.684 mhz sampling clock | 03:33 |
| lemay | gotcha - you already told me that, huh | 03:33 |
| lemay | :) | 03:33 |
| kristianpaul | oh no, sorry if i missudestand you, too many info and i'm soon (1rh or so) off bed | 03:34 |
| kristianpaul | you are in canada right? | 03:34 |
| lemay | so, since you're working with real samples, your next step is to try to correlate using osgps? | 03:34 |
| lemay | california | 03:34 |
| kristianpaul | oh, nice | 03:35 |
| kristianpaul | about osgps well not now, i did in the past some offline processing, as i was using milkymist just for data acquisition | 03:36 |
| kristianpaul | but now i have namuru on milkymist, so next plan is detect correlation peak | 03:37 |
| lemay | sounds good | 03:37 |
| kristianpaul | i hope ;) | 03:37 |
| lemay | does the namaru software take care of all of that? | 03:38 |
| kristianpaul | namuru is implemented in the fpga, lets said is the hardware DSP for all this, afaik no current software support for it that i'm aware off | 03:38 |
| kristianpaul | s/off/of | 03:39 |
| kristianpaul | so no no care of all that | 03:39 |
| kristianpaul | it will when the software side to drive the namuru will be developed/ported? for milkymist | 03:40 |
| kristianpaul | then we can call it milkymist gps baseband soc i guess ;-D | 03:40 |
| lemay | gotcha. So you're going to interface the FPGA on the milkymist with the RF front end - the SE4162T-EK1 - and use it to track GPS | 03:41 |
| lemay | this sounds plausible to me | 03:41 |
| kristianpaul | yes thats already done | 03:41 |
| kristianpaul | actually i hope from now that is missing is software side | 03:41 |
| lemay | Yea, I mean the whole thing - the tracking GPS part - that's plausible | 03:42 |
| lemay | what sort of CPU will you use? | 03:42 |
| wolfspraul | hey, Milkymist is a CPU :-) | 03:42 |
| kristianpaul | tell is a SoC.. | 03:42 |
| lemay | something in the milkymist - the LatticeMico32? | 03:42 |
| kristianpaul | milkymist uses a lattice mico32 | 03:42 |
| kristianpaul | yes | 03:43 |
| kristianpaul | but yes CPU is okay i think | 03:43 |
| lemay | haha, I don't know enough about the system architecture :) | 03:43 |
| lemay | well, I think it will work | 03:43 |
| kristianpaul | sure it will ! | 03:43 |
| lemay | I designed a front end for a system that is very similar | 03:44 |
| lemay | correlators in FPGAs, and controlled by an x86 running windows, of all things | 03:44 |
| lemay | its a Dell with a Virtex 4 | 03:44 |
| kristianpaul | s/tell/well | 03:45 |
| kristianpaul | wow wow, sounds fancy :) | 03:45 |
| kristianpaul | is it pusblished somwhere? | 03:45 |
| lemay | yea, its nifty | 03:45 |
| wolfspraul | lemay: can you share some sources? :-) | 03:45 |
| lemay | unfortunately I can't | 03:45 |
| lemay | but I can try to help you | 03:45 |
| kristianpaul | i'm open now to learn about pull-in algorythms | 03:46 |
| lemay | I understood the tracking loops pretty well at one point, but it will take me a bit to shake off the dust | 03:46 |
| kristianpaul | btw the software on the x86 worked upon accumlators? or implemented PLL directly on FPGA? | 03:46 |
| kristianpaul | i was told that was posible too | 03:47 |
| lemay | it worked on the early prompt lates from the fpga | 03:47 |
| kristianpaul | take your time lemay , and hang on here of course | 03:47 |
| kristianpaul | great, on namuru we have same :) | 03:47 |
| lemay | it updated the phase shift register based on those sums | 03:48 |
| lemay | and I suppose the code replicators, too | 03:48 |
| kristianpaul | code replicators? | 03:48 |
| lemay | yea - for the crrelation | 03:49 |
| kristianpaul | replicator = generator ? | 03:49 |
| kristianpaul | ah yeah, replica code, yes yes | 03:49 |
| kristianpaul | go ahead, sorry for interrupt | 03:49 |
| lemay | I've got to go afk for a few minutes | 03:50 |
| lemay | be back in 5 or so | 03:50 |
| kristianpaul | ok i think i'm going bed now, my head is not feeling good... | 03:52 |
| lemay | ok | 03:52 |
| lemay | nice talking to you | 03:52 |
| lemay | if you come up with any questions let me know | 03:53 |
| lemay | I am in #elphel a lot | 03:53 |
| kristianpaul | last question for you is about how you managed to your software be wideband and narrow band, | 03:53 |
| kristianpaul | also how to manage doppler, and when the receiver is moving | 03:53 |
| kristianpaul | okay, nice to meet you too lemay | 03:54 |
| lemay | Well, your sampling rate will allow you to adjust for doppler | 03:54 |
| kristianpaul | read you later | 03:54 |
| lemay | and you need to tune your tracking loops depending on what dynamics you want | 03:55 |
| kristianpaul | and how to manage that satellite moves... i mean i have this early promt late, but how i do correlate that data? is tha pull-in about? | 03:55 |
| kristianpaul | i really need to understand an algoythm for it.. | 03:55 |
| kristianpaul | yes i need to learn how in depth, at least to implement that for namuru, as i dont get that part clearly from osgps code... :| | 03:56 |
| lemay | the satellite motion is part of the doppler estimation | 03:56 |
| kristianpaul | ohh | 03:56 |
| kristianpaul | i tought was apart | 03:56 |
| lemay | I mean, that thing is moving at a jillion miles per hour, right? | 03:56 |
| lemay | so it has a doppler relative to you, even when your stationary | 03:57 |
| lemay | your tracking loops just care about the combined doppler | 03:57 |
| lemay | your combined doppler will be fairly insignificant compared with the correlation period | 03:57 |
| lemay | or, your delta doppler | 03:58 |
| lemay | I just started spouting gibberish, huh | 03:58 |
| lemay | so, you can tell what the combined doppler is, in general | 03:58 |
| kristianpaul | so whats the main issue about tracking? i mean once i detect the signal i play with... slewing code? or fixing local carrier nco values? | 03:59 |
| lemay | the main issue about tracking is replicating the correct code and carrier | 04:03 |
| lemay | you look at your early, prompt, and late correlation sums, and it tells you whether or not you are right on the signal, or if you are a little bit off | 04:04 |
| kristianpaul | i see, not sound complicated :) | 04:05 |
| lemay | based on that, you change the phase of the carrier that you are generating and you move the code forward or backwards | 04:05 |
| lemay | heh, are you being ironic or not? I can't tell. | 04:05 |
| kristianpaul | no no, no ironic | 04:05 |
| lemay | yea, conceptually it is not bad | 04:06 |
| kristianpaul | the faster the better i guess | 04:06 |
| lemay | so the PLL - phase locked loop - generates a carrier phase based on the early prompt and late sums (if I am remembering all this right.) | 04:06 |
| lemay | well, you have to correlate for so amount of time to generate your eraly prompt and late sums | 04:07 |
| lemay | like, 20 milliseconds is standard | 04:07 |
| kristianpaul | yes | 04:07 |
| lemay | so you get your updates at 50Hz | 04:07 |
| kristianpaul | yes, thats navigation rate per satellite right? | 04:08 |
| lemay | you update your tracking loops at 50hz | 04:08 |
| lemay | sounds right. | 04:08 |
| lemay | but thats different from the tracking loop | 04:08 |
| lemay | the tracking loop is adjusting the PLL at 50Hz, based on those correlation times | 04:09 |
| kristianpaul | tracking loop update in less than a 1ms i remenber, right? | 04:09 |
| kristianpaul | hum.. | 04:09 |
| lemay | the nav data is basically checking the sign of the prompt signal | 04:09 |
| lemay | well, depends on your implementation | 04:10 |
| lemay | if Namuru does 1ms that's great | 04:10 |
| lemay | and maybe that's normal, I haven't thought about this in a while - maybe tomorrow I'll be saying to myself, 'yea, duh, of course its 1 ms....' | 04:11 |
| kristianpaul | i think is less 1ms is too tight, but yes it does | 04:11 |
| kristianpaul | he sure, | 04:11 |
| lemay | so, narrowband vs wideband - I am not sure I understand the question | 04:12 |
| lemay | I'm pretty sure that we | 04:18 |
| lemay | heh, hit enter by accident :) | 04:19 |
| kristianpaul | mom | 04:23 |
| kristianpaul | http://www.datasheetcatalog.org/datasheet/zarlinksemiconductor/zarlink_GP2021_JAN_01.pdf page22 | 04:25 |
| kristianpaul | well,i said by now, off bed now | 04:28 |
| kristianpaul | chao :) | 04:28 |
| lemay | o/ | 04:28 |
| lemay | Yea, that sentence is kind of strange. I'd ignore it. | 04:28 |
| lemay | Essentially you change modes when you go from acquisition to tracking - in acquisition you are searching the entire code/doppler space, and once you have acquired, you don't need to search the doppler space - just the early prompt lates | 04:30 |
| rejon | here is a qi news item for you wolfspraul about shanzhai, blahblah http://www.iftf.org/node/3943 | 04:31 |
| rejon | most of my comments had to be removed because were too personal and stinging | 04:31 |
| rejon | esp. the one about the open hardware logo being a broken gear | 04:32 |
| wolfspraul | ha | 04:33 |
| wolfspraul | you are right, never thought about it | 04:34 |
| wolfspraul | broken gear :-) | 04:34 |
| wolfspraul | rejon: phew, so much text [iftf] - do I have to read all that? | 04:39 |
| wolfspraul | I scanned over Bunnie's comments and I share all of those views. | 04:40 |
| rejon | no | 04:40 |
| rejon | its pretty comprehensive look at shanzhai | 04:40 |
| wolfspraul | Bunnie does have a real and deep understanding of these things. | 04:40 |
| rejon | concludes that chapter | 04:40 |
| rejon | i think i just repeat most of the things you put in my head | 04:40 |
| rejon | but my comments were mostly removed by my request | 04:40 |
| rejon | i wanted that gear one to go in | 04:40 |
| rejon | anyway, next | 04:41 |
| rejon | just an item for the news | 04:41 |
| xiangfu | wpwrak, someone ask atBen in mailing list: (1) can it be connected to a Arduino microSD shield? I am using Digi | 07:42 |
| xiangfu | Xbees at the moment but I would rather switch to atben if possible | 07:42 |
| xiangfu | wpwrak, I don't know how Arduino software drivers works. so don't know how to answer :( | 07:43 |
| wpwrak | no, it won't work with the arduino microSD. arduino doesn't provide all the signals on the uSD. | 07:45 |
| wpwrak | and of course, you would also need to write a driver, etc. | 07:45 |
| xiangfu | ok. got it | 07:46 |
| qi-bot | [commit] Werner Almesberger: bitcmp/: little utility to find which bits differ between two files (master) http://qi-hw.com/p/wernermisc/08a13cd | 10:00 |
| qi-bot | [commit] Werner Almesberger: bitcmp/bitcmp.c: fixed fencepost error (master) http://qi-hw.com/p/wernermisc/78d9fda | 10:11 |
| `antonio` | wpwrak, I am trying to set up three nanonotes with atben, how do i set up the iz assoc and dirtpan? | 11:03 |
| wpwrak | iz assoc should be the same on the two non-coordinators (they will receive different addresses) | 11:05 |
| wpwrak | dirtpan is point-to-point, so you'll need two dirtpan instances on each node. | 11:05 |
| wpwrak | e.g., on the coordinator: | 11:05 |
| wpwrak | dirtpan 777 1 8001 'ifconfig $ITF $IP1 dstaddr $IP2 up' | 11:06 |
| wpwrak | dirtpan 777 1 8002 'ifconfig $ITF $IP1 dstaddr $IP2 up' | 11:06 |
| wpwrak | on the node with short address 8001: | 11:06 |
| wpwrak | dirtpan 777 8001 1 'ifconfig $ITF $IP2 dstaddr $IP1 up' | 11:06 |
| wpwrak | dirtpan 777 8001 8002 'ifconfig $ITF $IP2 dstaddr $IP1 up' | 11:06 |
| wpwrak | and on the other: | 11:06 |
| wpwrak | dirtpan 777 8002 1 'ifconfig $ITF $IP2 dstaddr $IP1 up' | 11:06 |
| wpwrak | dirtpan 777 8002 8001 'ifconfig $ITF $IP2 dstaddr $IP1 up' | 11:06 |
| wpwrak | (i hope :) have't tried three nodes myself yet | 11:07 |
| `antonio` | wpwrak, ok :) i'll let you know how that goes ! | 11:08 |
| jivs | Hi wpwrak, One of the atben is giving this error | 14:04 |
| jivs | i check the nanonote detects other atben, so i suppose kernel is fine | 14:04 |
| jivs | i tried on another nanonote as well same error | 14:04 |
| jivs | at86rf230 spi32766.0: Non-Atmel device found (MAN_IDff ff) | 14:04 |
| jivs | [ 3.240000] at86rf230: probe of spi32766.0 failed with error -22 | 14:04 |
| jivs | what it could be? anything related to atben hw? | 14:04 |
| `antonio` | wpwrak, 3 nanonote connected together with atben, that worked, thanks | 14:43 |
| wolfspraul | wow | 14:44 |
| wolfspraul | `antonio`: three? | 14:44 |
| `antonio` | yes | 14:45 |
| Ayla | interesting | 14:46 |
| jivs | 4 of them now! | 14:55 |
| wpwrak | jivs: (error) sounds like the atben card is missing or makes bad contact | 15:28 |
| wpwrak | whee ! largest atben network on the planet ! :) | 15:29 |
| Action: kristianpaul feels bad not buying/supporting some atben.. | 15:30 | |
| wpwrak | traitor !!! :) | 15:30 |
| jivs | wpwrak, i tried few times and on few ben nanonotes.. | 15:31 |
| jivs | but i will try again if any device can detect it.. | 15:32 |
| kristianpaul | no rush with tuxbrain. but paying lot of taxes make me keep away :( | 15:36 |
| kristianpaul | may be sharism can sell me soem atben? :) | 15:37 |
| kristianpaul | `antonio`: what are your plans with the atben? | 15:40 |
| `antonio` | kristianpaul, I am testing atbens as part of a project and hopefully tomorrow i'll do a video | 15:44 |
| `antonio` | wpwrak, for that jivs error: one atben is faulty. | 15:45 |
| wpwrak | `antonio`: :-( | 15:47 |
| wpwrak | `antonio`: if you want to analyze what's wrong, you could run the production test software on it | 15:49 |
| wpwrak | `antonio`: most likely, the GPIO test would spot something | 15:49 |
| wpwrak | `antonio`: the test is in this script: ben-wpan/prod/atben | 15:55 |
| `antonio` | ok | 15:57 |
| `antonio` | i'll let you know how that goes | 15:58 |
| wpwrak | there's actually very little that can go wrong on atben. any failure wold have to be a short (foreign particle ?) or a solder joint that failed | 15:59 |
| wpwrak | for failed solder joint, only a joint on the transceiver itself or the crystal would make it become unresponsive | 16:00 |
| wpwrak | for a short, add the capacitors as candidates | 16:00 |
| `antonio` | interesting ! | 16:05 |
| wpwrak | `antonio`: setup and usage instructions for the test system are here: http://downloads.qi-hardware.com/people/werner/wpan/prod/ | 16:21 |
| wpwrak | `antonio`: in your case, you don't need the atusb side (i.e., later tests of the atben would fail without atusb to talk to around) | 16:21 |
| `antonio` | wpwrak, i went through that already last week and I had some problem, i'll let you know if I got that working | 16:27 |
| `antonio` | also because i'll need your help :) | 16:29 |
| wpwrak | hehe :) | 16:30 |
| qi-bot | [commit] Werner Almesberger: bitcmp/bitcmp.c: added indication of direction of bit flip (1 -> 0 or 0 -> 1) (master) http://qi-hw.com/p/wernermisc/dd0c4ec | 16:36 |
| methril_work | wpwrak, in your fw adventures do you ever have stack problems? | 16:43 |
| wpwrak | methril_work: so far, i never hit any. at least not that i know ;-) | 16:45 |
| methril_work | wpwrak, i`m in this nightmare at work... | 16:46 |
| methril_work | i know because a wonderful processor feature that resets the preocessor if a StackOverflow occurs | 16:46 |
| wpwrak | can you disable the reset ? (-:C | 16:48 |
| methril_work | yes, but then it gives undetermined behaviour | 16:49 |
| methril_work | it`s worst disabling it | 16:49 |
| wpwrak | excellent. then you can blame the hardware :) | 16:50 |
| methril_work | i balme all the time!! :) | 16:50 |
| methril_work | s/balme/blame/ | 16:50 |
| wpwrak | :) | 16:50 |
| wpwrak | in hardware we distrust :) | 16:50 |
| methril_work | not in this HW :) | 16:51 |
| kristianpaul | arggg cannot exec `cc1' i hate rpm NOW | 17:21 |
| kristianpaul | what's license for this mmtp_i2l.pdf btw¿ | 18:27 |
| kristianpaul | s/¿/? | 18:27 |
| lekernel | wtfpl | 19:50 |
| kristianpaul | no other license please | 20:09 |
| DocScrutinizer | yay exception on stack overflow - cool feature | 20:13 |
| wpwrak | PIC, i guess ? :) | 20:15 |
| lekernel | actually I wonder why no other arch implements this SPLIM register | 20:16 |
| lekernel | it's not even particularly hard to do | 20:16 |
| lekernel | we should add that to lm32 :) | 20:17 |
| wpwrak | lekernel: why ? the MMU will catch it nicely ;-) | 20:17 |
| lekernel | SPLIM is much easier to support in RTEMS than a MMU-based system | 20:18 |
| lekernel | and it will take ages before Flickernoise runs flawlessly on Linux | 20:20 |
| Action: wpwrak happily notes the now the possibility of running flickernoise on linux is being considered :) | 20:21 | |
| wpwrak | but yes, certainly nothing trivial | 20:21 |
| lekernel | I have nothing against Linux... except that it doesn't work, and getting it to work takes 50-100x as much time as RTEMS for equivalent tasks | 20:23 |
| wpwrak | but when it works ... :) | 20:24 |
| kristianpaul | people will blame the 80Mhz cpu.. | 20:26 |
| wpwrak | blame or dislike ? | 20:26 |
| wpwrak | isn't there still quite a bit of spare room in the fpga ? maybe you could add more cores ;-) | 20:27 |
| kristianpaul | yes, sorry. dislike | 20:27 |
| kristianpaul | oh sure | 20:27 |
| kristianpaul | all is posible ! ;) | 20:27 |
| wpwrak | simplify it and make a decacore. more than all the other big players ;-) | 20:28 |
| lekernel | in fact, a decacore might work right now | 20:29 |
| wpwrak | add some branch elimination logic for the bogomips loop, and you'll have stellar values there, too ;-) | 20:29 |
| lekernel | except that there's no software support for it ofc | 20:29 |
| wpwrak | lekernel: who cares about sw support ? it's all marketing :) | 20:29 |
| kristianpaul | just lack software support, what about hdl part? is that easy as just adding another lm32 to the shared bus? | 20:31 |
| wpwrak | (branch elimination) when you take a branch, remember where it went and what the condition was. then have a comparator on the program counter that implicitly executes the branch when incrementing into the address of the branch instruction iff the condition is met. voila, removed at least 1 instruction cycle :-) | 20:33 |
| wpwrak | now quick, where do i file the patent :) | 20:33 |
| lekernel | kristianpaul, if you don't want cache coherency (which could be a problem depending on the software) yes | 20:34 |
| lekernel | if you want cache coherency... you have to dig into the CPU pipeline, and since people seem already afraid of the simpler MMU this has little chance of happening | 20:34 |
| lekernel | though the LM32 cache is write-through, so all you have to do is broadcast writes to all cores. not too hard. | 20:36 |
| wpwrak | i somehow feel that we may see more cores pop up before too long :) | 20:37 |
| lekernel | wpwrak, phew, I'm not that optimistic | 22:06 |
| lekernel | among the current people I don't think anyone would pull that off | 22:07 |
| lekernel | and wrt finding new people, my google analytics reports still feel like a spit in my face | 22:07 |
| lekernel | so... it will be hard | 22:07 |
| wpwrak | lekernel: ah, i was thinking that you might have had an inspiration of the "hey, this would be EASY" kind :) | 22:08 |
| wpwrak | (followed by a few hours of furious hacking. then a few more days, realizing that it wasn't all that EASY, but still doable. etc. ;-) | 22:10 |
| lekernel | can you believe it? the elektor article, run in 3 different languages (and soon 4), generated so far exactly 2 people following the project and 1 board sale | 22:12 |
| wpwrak | wow. does elektor have so little reach these days ? | 22:15 |
| lekernel | seems so | 22:15 |
| lekernel | even slashdot had better results (and it took less time to write the article, too), despite the trolling | 22:15 |
| wpwrak | heh :) | 22:16 |
| wpwrak | well, a good article is also something you can point people to and you can reuse some of the material. so it's not a waste of time | 22:16 |
| lekernel | so atm i'm trying xcell, aimed specifically at people who are not afraid of fpga's to start with | 22:17 |
| lekernel | some articles are also "syndicated" in publications like eetimes | 22:18 |
| wpwrak | yeah. maybe that'll get some more developers aboard | 22:18 |
| lekernel | so, atm I rather feel the need to fix that huge unpopularity problem rather than mess with multicore lm32's ... | 22:19 |
| wpwrak | maybe multicore will get people excited :) | 22:24 |
| wpwrak | i think once linux runs well, this may help. lowers the barrier of entry quite a bit if you have a familiar operating system. | 22:25 |
| --- Thu Aug 18 2011 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!
