| 00:01.05 | ka6sox-office | vmaster, thanks! |
| 00:01.22 | beewoolie-afk | I found one goofy thing in the ftdi protocol. |
| 00:01.31 | ka6sox-office | oh? |
| 00:01.40 | ka6sox-office | only 1? |
| 00:01.45 | beewoolie-afk | It's the sort of thing that makes this not such an attractive style of handle the transfer. |
| 00:01.48 | beewoolie-afk | 1 so far. |
| 00:02.20 | beewoolie-afk | When writing data to a register, the last bit should be written with TMS=1 so that the TAP leave the SHIFT state. |
| 00:02.51 | beewoolie-afk | The way that the ftdi handles this is they leave off the last bit and use another command to shift that last bit while also holding TMS high. |
| 00:03.01 | ka6sox-office | weird |
| 00:03.32 | beewoolie-afk | I can see why they did it that way, but it betrays the underlying sorts of problem we'll see when putting lots of logic in the FPGA. |
| 00:04.13 | beewoolie-afk | So here's the essence of the trade off. |
| 00:04.28 | beewoolie-afk | The idea I'd been playing with was to use a byte per clock transition. |
| 00:04.46 | beewoolie-afk | That means that a scan for 600 cells in a scan chain would be about 610 bytes. |
| 00:04.56 | beewoolie-afk | Not a big deal when we are in the same address space. |
| 00:05.22 | beewoolie-afk | The same operation using an ftdi protocol is about 80 bytes. |
| 00:05.45 | beewoolie-afk | I'm going to spec both and see what people think. |
| 00:05.52 | ka6sox-office | sounds good. |
| 00:05.56 | beewoolie-afk | We *can* support both without much grief. |
| 00:17.19 | lennert | beewoolie-afk: what i proposed to ka6sox earlier was a 'scan finished' bit (when you scan to tdi) which would toggle tms on the last clock to leave shift-{dr,ir} state |
| 00:17.48 | beewoolie-afk | I don't know what you mean. |
| 00:18.36 | beewoolie-afk | I'm going to code for a IO opcode that has a bit which, when set, will latch TMS=1 on the last data bit. |
| 00:23.31 | beewoolie-afk | ka6sox-office: How much RAM is in one of these FPGAs? Assuming that I want it to store some data locally. |
| 00:25.05 | AchiestDragon | the XC3S400-4TQ144C has about 200k bits |
| 00:25.45 | beewoolie-afk | 25KiB. |
| 00:27.03 | AchiestDragon | actualy 36Kbytes 288kbits |
| 00:29.40 | beewoolie-afk | lennert: Actually, we can do one better. We can *always* transition out of the shift state after an IO operation and then reenter it if we need to. |
| 00:31.50 | lennert | beewoolie-afk: sure, that sounds sane.. when scanning to tdi, always set tms=1 on the last bit |
| 00:32.09 | beewoolie-afk | Doesn't matter, TDI or TDO. |
| 00:32.19 | beewoolie-afk | It's safe for either. |
| 00:32.29 | lennert | beewoolie-afk: you could have an extra bit for 'leave scan-{dr,ir} when i am done' but you could also just do it anyway, indeed |
| 00:32.31 | beewoolie-afk | The ftdi protocol is generalized for more than just JTAG. |
| 00:32.36 | lennert | well, TDI/TDO i regard as one case |
| 00:32.47 | lennert | i mean, inward scans go either to tms or tdi |
| 00:32.50 | beewoolie-afk | I had a bit at first, but I think now it is saner to do it this way. |
| 00:33.35 | beewoolie-afk | Yeah, though TMS scanning isn't necessary because I'm using a SeekState command to put the TAP into the next desired state. |
| 00:33.54 | lennert | beewoolie-afk: document your protocol and i'll read it |
| 00:34.00 | beewoolie-afk | There is only one place where it may be useful to use TMS scanning. |
| 00:34.16 | beewoolie-afk | At least, I've only seen one. |
| 00:34.17 | lennert | okay |
| 00:36.39 | ka6sox-office | beewoolie-afk, we also can put a 32MB SDRAM on the board too. |
| 00:37.18 | beewoolie-afk | Interesting idea. If the transactions are of differing lengths, is the FPGA going to be able to keep lists of them efficiently? |
| 00:40.23 | AchiestDragon | it should could have a bit counter so it sould know how many bits to send and how much to/has recived |
| 00:40.50 | beewoolie-afk | AchiestDragon: For you, I'll put in three. That'll give us two spares. :-) |
| 00:46.00 | beewoolie-afk | I'm including ways to manipulate the reset signals. Do we need to have ways to operate some extra GPIO lines? |
| 00:48.35 | beewoolie-afk | ka6sox-office: I know that the ep1220 board has some GPIO lines connected to the JTAG socket. Are any of these things we're going to need to manipulate? |
| 00:49.44 | key2 | hi |
| 00:49.54 | beewoolie-afk | wb |
| 00:51.28 | ka6sox-office | beewoolie-afk, probably |
| 00:51.37 | beewoolie-afk | Can you be more specific? |
| 00:51.45 | ka6sox-office | I would have to look at the schematic. |
| 00:51.57 | key2 | someone here has ever used Mentor Graphic to make a layout for a PCB ? |
| 00:52.06 | beewoolie-afk | Well, OK. From that point of view, perhaps I can make something make sense. |
| 00:53.26 | ka6sox-office | key2, I haven't |
| 01:10.58 | [g2] | beewoolie-afk lennert ping |
| 01:11.04 | beewoolie-afk | [g2]: hi |
| 01:11.10 | [g2] | hey |
| 01:11.36 | [g2] | Ok Gatworks has a FPGA based parallel dongle that'll do JTAG |
| 01:11.55 | [g2] | TCLK runs at 8MHz |
| 01:12.06 | [g2] | and it is driven over the parallel port |
| 01:12.12 | ka6sox-office | about 5X too slow |
| 01:12.21 | beewoolie-afk | Still an interesting device, tho. |
| 01:12.27 | ka6sox-office | yep |
| 01:12.29 | [g2] | ka6sox-office IXP only does 10MHz |
| 01:12.29 | beewoolie-afk | They must have some sort of protocol to handle it. |
| 01:12.43 | [g2] | no protocol |
| 01:12.54 | [g2] | it's a serial to parallel issue |
| 01:13.00 | ka6sox-office | but a proprietary driver? |
| 01:13.23 | [g2] | It's a very simple interface we'll get an API |
| 01:13.38 | [g2] | I'm gonna have 5 devices shipped to me tomorrow |
| 01:13.40 | beewoolie-afk | There has to be something because I don't know how they'd clock the parallel port that fast. |
| 01:14.05 | [g2] | they don't clock the parallel port that fast, they clock the FPGA that fast |
| 01:14.32 | beewoolie-afk | so then they have to send commands to the FPGA. Hence, a protocol. |
| 01:15.38 | [g2] | we'll they'll write up the interface and send it to me soon. They said they can drive the chip at 8MHz which is fast enough for my dev kit and nearly as fast as we can drive the XScales |
| 01:16.16 | [g2] | So for IXP425 and probably PXA27x it's 80% line rate |
| 01:16.42 | [g2] | I'll have to see the specs for parallel port mode they are running in. |
| 01:17.05 | [g2] | what is it EPP or ECP and bi-directional mode etc.... |
| 01:17.29 | beewoolie-afk | According to the BSDL file, theIXP can be clocked at 16MHz. |
| 01:18.51 | *** join/#openjtag vmaster_ (i=vmaster@p549B5463.dip.t-dialin.net) |
| 01:21.48 | [g2] | beewoolie-afk I think we could also reload the FPGA to make it do whatever we wanted |
| 01:22.07 | beewoolie-afk | Do you mean that we could reload the FPGA on these Gateworks devices? |
| 01:22.09 | AchiestDragon | there are problems driving a parallel port a high clocking rates in linux , even rt linux , dos is faster but its still slow in real terms , the guys in #emc should know the rates posible , as its used for driving steppers for cnc |
| 01:22.31 | [g2] | beewoolie-afk I think that is probably an option |
| 01:22.45 | beewoolie-afk | That would be an interesting test apparatus. |
| 01:23.00 | [g2] | the current driver runs in dos mode |
| 01:23.52 | [g2] | Hey I'm learning a little about real time linux or fine with just turning the interrtups off for several minutes |
| 01:23.55 | AchiestDragon | linux is multitasking where dos is not , so dos programs tend to run a bit faster |
| 01:24.32 | [g2] | put the screen in VGA mode and write to the buffer in BIOS |
| 01:25.12 | [g2] | or maybe everyone needs an X2 AMD64 to drive the thing :) |
| 01:25.31 | [g2] | dedicate on processor to servicing he parallel port :) |
| 01:25.37 | [g2] | s/on/one/ |
| 01:26.47 | AchiestDragon | or ditch using an os and write a program that does what you want in asm from boot up |
| 01:34.40 | [g2] | AchiestDragon just patch grub :) |
| 01:34.48 | [g2] | add a "command" |
| 01:35.01 | [g2] | boot Debain, |
| 01:35.07 | [g2] | boot Ubuntu |
| 01:35.12 | [g2] | boot Gentoo |
| 01:35.22 | [g2] | boot Bad as JTAG programmer ? |
| 01:35.30 | [g2] | s/as/ass/ |
| 01:36.15 | beewoolie-afk | Yeah. Real bad as. |
| 01:36.17 | [g2] | AchiestDragon one of the guys #emc thinks the parallel port can do 400kBs in ECP mode |
| 01:36.40 | AchiestDragon | was thinking more on lines of write native code in asm to directly drive the parallel port and other i/o as a single program , then just flash it into the bios of the mobo |
| 01:37.00 | beewoolie-afk | AchiestDragon: that is bad as. |
| 01:37.11 | beewoolie-afk | s/as/az/ |
| 01:37.18 | AchiestDragon | power on and it does the job |
| 01:37.34 | beewoolie-afk | Would be fine for an old 386. Maybe? |
| 01:37.47 | [g2] | except for where you get the data from :) ? |
| 01:38.04 | beewoolie-afk | CF card. |
| 01:38.20 | [g2] | right like lots of BIOS read CF cards |
| 01:38.31 | beewoolie-afk | CF cards are really easy to read. |
| 01:38.40 | AchiestDragon | or floppy , if you write floppy drivers to read an image |
| 01:38.47 | beewoolie-afk | He's going to write a n ASM program. |
| 01:38.56 | beewoolie-afk | Sure. Floppy is about the same complexity. |
| 01:39.04 | AchiestDragon | hdd if you want to write the file system drivers |
| 01:39.05 | [g2] | CF much better |
| 01:39.22 | [g2] | same problem on CF or IDE |
| 01:39.26 | AchiestDragon | or usb memory stick |
| 01:39.45 | AchiestDragon | and same problem there |
| 01:40.06 | [g2] | well the CF/IDE interface is really simple and can be done in PIO mode |
| 01:40.17 | AchiestDragon | unless you stored the data in a raw sector format |
| 01:40.40 | beewoolie-afk | AchiestDragon: How much time do you have? |
| 01:40.55 | AchiestDragon | not enough to do it that way |
| 01:41.07 | [g2] | beewoolie-afk the question is how l337 are you ? |
| 01:41.28 | beewoolie-afk | I'm not young enough to be l337 |
| 01:42.04 | AchiestDragon | is a aging geek here 41 |
| 01:42.56 | beewoolie-afk | AchiestDragon: you're in like company |
| 01:42.57 | [g2] | kids these days |
| 01:44.34 | AchiestDragon | yes , there lucky , they can actualy buy pre built computers now , rather than having to build them from kits |
| 01:44.56 | [g2] | yeah that's why they don't understand them :) |
| 01:49.03 | [g2] | beewoolie-afk you've got C code to drive the parallel port now ? |
| 01:49.28 | beewoolie-afk | [g2]: I've been using ppdev, but I'll probably write something to use the port directly, too. |
| 01:49.53 | [g2] | do you know what mode ppdev uses ? |
| 01:49.55 | beewoolie-afk | I'm done with the first draft of the protocol. I'll review it and see about posting it to the wiki. |
| 01:50.52 | beewoolie-afk | [g2]: My laptop BIOS sets the port in SP mode. I think that's what it's called. |
| 01:53.13 | beewoolie-afk | SPP. |
| 02:04.54 | *** join/#openjtag bullet (n=bullet@90.55.62.81.cust.bluewin.ch) |
| 02:45.33 | wookey__ | I can't keep up with this channel! |
| 02:45.44 | ka6sox-office | wookey__, why? |
| 02:45.50 | wookey__ | Is someone collecting all the useful URL's that are going by |
| 02:46.03 | wookey__ | ka6sox-office: becauswe I have other things to do too |
| 02:46.13 | ka6sox-office | they are in the log and I'm trying to get them in the wiki |
| 02:46.15 | wookey__ | apart from read IRC |
| 02:46.24 | ka6sox-office | www.openjtag.net |
| 02:46.51 | wookey__ | OK. I looked over g2's pdf summary, but that's all I've had time to read |
| 02:47.05 | ka6sox-office | k |
| 02:47.14 | ka6sox-office | how close is Balloon V3? |
| 02:47.25 | wookey__ | didn't you just ask me that? |
| 02:47.44 | ka6sox-office | yes but i didn't see an answer. |
| 02:47.51 | wookey__ | oh, I sent some |
| 02:48.00 | wookey__ | oh - are we on freenode here? |
| 02:48.06 | ka6sox-office | yes |
| 02:48.07 | wookey__ | that won't let me PM |
| 02:48.17 | ka6sox-office | and your nick is probably not registered... |
| 02:48.28 | wookey__ | becauae my nicks not registerded (because someone else has got it and variants) |
| 02:48.35 | ka6sox-office | ya |
| 02:48.37 | ka6sox-office | np |
| 02:48.54 | ByronT | and I just thought I was getting ignored all the time..... |
| 02:48.57 | wookey__ | I don't like freenode (I get kicked off when they other 'wookey' turns up too |
| 02:48.58 | [g2] | wookey__ feel free to e-mail comments to me |
| 02:49.34 | wookey__ | anyway - yes SFAIK it's on schedule - boards done and build scheduled for a week or two's time |
| 02:49.46 | wookey__ | but I'm not in the loop and there may be delays no-one has told me about |
| 02:50.07 | ka6sox-office | hokey dokey |
| 02:50.09 | ka6sox-office | thanks |
| 02:50.20 | wookey__ | (last balloon2 build is delayed due to too many failures on the boards, which is odd, because they have done plenty of batches before |
| 02:51.46 | ka6sox-office | bummer |
| 02:54.11 | ka6sox-office | nite |
| 02:54.48 | ByronT | I keep trying to tell myself that sleep is overrated, but for some reason, I don't listen |
| 02:55.18 | ka6sox-office | heh |
| 03:10.20 | beewoolie-afk | ka6sox-office: OK. Ready for v1.0? |
| 03:10.40 | beewoolie-afk | I think I'll let you do the wiki magic. |
| 03:10.50 | beewoolie-afk | ftp://ftp.buici.com/pub/jtag/atp.txt |
| 03:13.43 | ka6sox-office | thanks |
| 03:13.47 | ka6sox-office | I'll publish it :) |
| 03:14.21 | beewoolie-afk | I've understood the ftdi protocol adequately and I think I've got everything that they do that we care about. |
| 03:14.40 | beewoolie-afk | I'm still going to look over vmaster's comments again to see how his ideas fit in. |
| 03:15.06 | ka6sox-office | okay this looks pretty easy to put in the wiki. |
| 03:15.20 | beewoolie-afk | As long as the boxes stay aligned. |
| 03:15.27 | ka6sox-office | I'll work on that. |
| 03:18.58 | beewoolie-afk | One of the ways that ATP betters ftdi is that it has a command to put the TAP in a given state. |
| 03:19.34 | beewoolie-afk | While this doesn't significantly reduce the size of messages, it does make it easier to knit them together. When, for example, we start using macros. |
| 03:19.43 | ka6sox-office | k |
| 03:20.48 | beewoolie-afk | The way I was previously thinking of doing this was just the TRANSACT opcode. |
| 03:21.05 | beewoolie-afk | It explicitly sets the values of each control line. |
| 03:21.17 | beewoolie-afk | The BULKIO is more the way that the ftdi handles things. |
| 03:23.04 | beewoolie-afk | I hesitate to put information about each of the cores on the chain into this command protcol. The same thing can be done by loading commands into the FPGA with the right numbers of bits. |
| 03:23.40 | beewoolie-afk | Once the controller reads the IDCODES, it knows how many cores there are. It can either use BSDL files, or probing to determine the length of each scan chain. |
| 03:24.19 | beewoolie-afk | I believe that the controling software needs to know the length and encoding of the instruction register as these numbers aren't standardized. |
| 03:25.22 | beewoolie-afk | The first place I'd optimize is in polling status bits on scan chains. |
| 03:26.16 | beewoolie-afk | The ARMv4 has a feature where the host needs to poll for a bit in a scanchain. When that bit (bits) has the right state, it knows that the function is complete. |
| 03:26.44 | beewoolie-afk | I can see adding a OPCODE to read, compare, repeat, until the value is detected, or there is a timeout. |
| 03:27.38 | beewoolie-afk | A more interesting case is the DCC IO registers. Both the ARMv4 and the XSCALE have these. |
| 03:28.31 | beewoolie-afk | We could get some really great throughput by allowing an opcode to read and/or write the DCC values in the FPGA since it can check the status bits very efficiently. |
| 03:29.19 | beewoolie-afk | Note that this requires, in both cases, a monitor program on the target that can handle the other side of the exchange. |
| 03:39.11 | ka6sox-office | look at the wiki here and tell me if its okay: http://www.openjtag.net/wiki/Info/AcceleratedJTAG |
| 03:44.39 | ka6sox-office | Time to go home...back in a while from there. |
| 04:02.22 | beewoolie-afk | Looks fine. |
| 04:02.27 | beewoolie-afk | I'll add a few more comments. |
| 04:35.22 | *** join/#openjtag ka6sox (n=ka6sox@nslu2-linux/ka6sox) |
| 04:47.08 | ka6sox | beewoolie-afk, did you have a chance to look at the wiki? |
| 04:47.25 | beewoolie-afk | 20:00 |
| 04:47.33 | ka6sox | thanks. |
| 04:47.36 | beewoolie-afk | I even updated it. |
| 04:47.38 | ka6sox | I was "afk" |
| 04:47.40 | ka6sox | cool! |
| 04:47.47 | ka6sox | okay its okay with you then. |
| 04:48.17 | ka6sox | now everyone can look and comment as well as create other supporting documents. |
| 04:48.23 | beewoolie-afk | Well , I didn't read through all of it, but I assume that you copied it verbatim. |
| 04:48.38 | ka6sox | cut and paste verbatim |
| 04:49.39 | ka6sox | I need to go back over the logs and put the urls that we have been discussing in the wiki. |
| 04:49.52 | ka6sox | unless somebody else feels like they want to. |
| 04:50.02 | beewoolie-afk | so, while I know what vmaster would like to add to the protocol, I'm not certain what will make sense as a general purpose extension. |
| 04:50.12 | beewoolie-afk | knock yourself out. |
| 04:50.48 | ka6sox | I was hoping to find a volunteer as I'm really needing to populate a new server for the project. |
| 04:51.01 | beewoolie-afk | for this project? |
| 04:51.10 | ka6sox | and others. |
| 04:51.34 | ka6sox | this project is getting a new server that will also serve up CVS. |
| 04:51.35 | beewoolie-afk | Part of my reluctance to get too involved with the wiki is that I don't like the formatting. |
| 04:51.51 | ka6sox | I think that most of us will be working on different sections. |
| 04:51.52 | beewoolie-afk | Can we go to SVN. CVS is so last century. |
| 04:52.02 | ka6sox | security on SVN sucks. |
| 04:52.12 | ka6sox | its either all or nothing. |
| 04:52.14 | beewoolie-afk | My objection is with this particular one. |
| 04:52.20 | beewoolie-afk | SVN is fine if you use ssh. |
| 04:52.42 | beewoolie-afk | CVS sucks because it has no security. |
| 04:52.49 | beewoolie-afk | And it's kinda dated. |
| 04:53.16 | beewoolie-afk | Actually, I'd really like to move to git, but I've been waiting for the churn to slow down. |
| 04:53.21 | ka6sox | we have an application called lysCVS that handles the security issues. |
| 04:53.24 | beewoolie-afk | It's really fast and it makes merges really easy. |
| 04:53.32 | ka6sox | okay |
| 04:53.41 | ka6sox | is Git ready for prime time? |
| 04:53.41 | beewoolie-afk | Are you using :pserver:? |
| 04:53.51 | ka6sox | I have :pserver: tool. |
| 04:53.55 | ka6sox | er too |
| 04:53.56 | beewoolie-afk | As far as I can tel, git is good to go. They've released 1.1 |
| 04:54.04 | beewoolie-afk | :pserver: is really weak security wise. |
| 04:54.10 | beewoolie-afk | Passwords are sent as clear-texgt. |
| 04:54.13 | ka6sox | ugh |
| 04:54.15 | beewoolie-afk | s/gt/t/ |
| 04:54.25 | ka6sox | I use keys |
| 04:54.33 | ka6sox | developer keys |
| 04:54.42 | beewoolie-afk | :pserver: with keys? I'm not aware of such a thing. |
| 04:54.51 | beewoolie-afk | Perhaps that isn't :pserver:. |
| 04:54.54 | ka6sox | just like sourceforge. |
| 04:55.22 | ka6sox | anonymous can only download to upload you send me your keys and I add you. |
| 04:55.25 | beewoolie-afk | Well, I've not tried to share with SVN. As a tool, it makes CVS look like a tricycle. |
| 04:55.35 | ka6sox | :) |
| 04:56.02 | ka6sox | I think that most of us will be working on different areas so overlap (except at the edges) should be minimal. |
| 04:56.13 | beewoolie-afk | it branches and keeps tags really well. However, I think tha GIT is a better solution all around. It solves the merging problem which neither CVS nor SVN do very well. |
| 04:56.16 | ka6sox | but I'm willing to investigate other CMS. |
| 04:56.24 | beewoolie-afk | I get that. |
| 04:56.25 | AchiestDragon | pmsl ,, running tts on chat here , the filters for emiticons tanslates :p to " sticks out tung " so :pserver: is getting spoken as " sticks out tung server " |
| 04:56.34 | ka6sox | Monotone is waaaay too slow. |
| 04:56.35 | beewoolie-afk | I don't put my repositories on the web at the moment. |
| 04:56.41 | beewoolie-afk | Not going there. |
| 04:56.54 | beewoolie-afk | I've see the hassles from the openslug group. |
| 04:57.07 | beewoolie-afk | Git is amazingly fast. |
| 04:57.14 | beewoolie-afk | SVN is good, but git is better. |
| 04:57.16 | ka6sox | how is the security? |
| 04:57.30 | beewoolie-afk | It doesn't have security issues because patches have to be applied. |
| 04:57.36 | beewoolie-afk | Everyone has a repo. |
| 04:57.44 | beewoolie-afk | It's the way the kernel works. |
| 04:57.54 | beewoolie-afk | No one can 'checkin' without passing a patch. |
| 04:58.09 | beewoolie-afk | There may be a way to push changes, but I haven't seen how it works. |
| 04:58.27 | beewoolie-afk | It is possible that there is a way to 'trust' users and have patchsets be pulled automatically. |
| 04:59.04 | ka6sox | in this case we aren't making patches. |
| 04:59.09 | beewoolie-afk | I've been slow to get too deep since I don't need another tool to learn. I got enough of bitkeeper working to be competent, but it was poop. |
| 04:59.21 | beewoolie-afk | any change to the repo is a patch. |
| 04:59.42 | ka6sox | for the kernel or this project? |
| 04:59.55 | ka6sox | the kernel certiainly. |
| 04:59.58 | beewoolie-afk | well, any project. It's just a model of development. |
| 05:00.03 | ka6sox | k |
| 05:00.25 | beewoolie-afk | any new code can converted into a patch against a previous version of the project. |
| 05:00.44 | beewoolie-afk | GIT models projects as a tree of files. |
| 05:00.57 | beewoolie-afk | There are 'porcelain' programs that make it look like what people are used to. |
| 05:01.20 | ka6sox | I see that. |
| 05:01.34 | ka6sox | Core GIT, Cogito, StGIT, (h)gct, gitk, qgit, or gitweb? |
| 05:01.56 | ka6sox | too many choices and they all I'm sure are good at somethings but not others. |
| 05:02.08 | beewoolie-afk | Actually, there is more order than you might expect. |
| 05:02.18 | ka6sox | sooo...I also dont' have a lot of time to investigate this. |
| 05:02.18 | beewoolie-afk | Cogito is the most interesting. |
| 05:02.25 | ka6sox | I've heard of that. |
| 05:02.33 | beewoolie-afk | Others are there for funny other kinds of functionality. |
| 05:02.48 | beewoolie-afk | Git can be used standalone, but cg makes it quite simple. |
| 05:03.01 | beewoolie-afk | ka6sox: I'm going to have to learn it some time since I have patches for the kernel. |
| 05:03.10 | ka6sox | k |
| 05:03.16 | beewoolie-afk | At that point, I'll be in a better position to deal with comparing it to other things. |
| 05:03.26 | ka6sox | cool |
| 05:03.37 | beewoolie-afk | Certainly, CVS is fine for the time being. Nearly everyone knows it. |
| 05:03.43 | ka6sox | yep |
| 05:03.57 | beewoolie-afk | If SVN were a slam dunk, I'd push for it. But I'm not that concerned. |
| 05:03.58 | ka6sox | came with the training wheels. |
| 05:04.29 | ka6sox | if you have multiple projects a dev in one project *can* make changes to another repo. |
| 05:04.38 | ka6sox | and can mess things up. |
| 05:04.53 | ka6sox | (not quite as bad as MTN but almost!) |
| 05:05.27 | beewoolie-afk | MTN? |
| 05:05.38 | beewoolie-afk | Mack The Nife? |
| 05:06.10 | ka6sox | Monotone |
| 05:06.15 | beewoolie-afk | Oh my. |
| 05:06.25 | beewoolie-afk | I tried mercury, too. |
| 05:06.31 | ka6sox | to really mess things up it takes Monotone :) |
| 05:06.43 | beewoolie-afk | Frankly, and I know it's bad for 'someone's |
| 05:06.45 | beewoolie-afk | <PROTECTED> |
| 05:06.56 | beewoolie-afk | But, Linus really does get a lot of things right. |
| 05:07.08 | beewoolie-afk | GIT really is a very good idea. |
| 05:08.07 | ka6sox | I'll consider it when I have a little time. |
| 05:10.36 | ka6sox | beewoolie-afk, do you have a TS7200? |
| 05:11.13 | beewoolie-afk | you're kidding, right? |
| 05:12.16 | ka6sox | he he |
| 05:12.52 | beewoolie-afk | I've got stacks of arm7 and arm9 based boards. |
| 05:12.56 | beewoolie-afk | No ts7200. |
| 05:13.06 | ka6sox | okay anything with PC104 on it? |
| 05:21.16 | beewoolie-afk | ka6sox: I used to have a 386 board that was PC104. I don't remember what was up with it. I think it's got a broken component, so it was trash to the client. It worked as far as I can recall. |
| 05:23.03 | ka6sox | cool |
| 05:30.23 | ka6sox | I have the PC104 spec and the ISA timing diagrams so I can work with that for the VHDL to make the bus work. |
| 05:36.28 | [g2] | beewoolie-afk Nice write up... |
| 05:36.43 | beewoolie-afk | [g2]: thx/ |
| 05:37.16 | beewoolie-afk | I like having something concrete to reference. It makes it easier to say factually what will happen. |
| 05:37.17 | [g2] | I think we should figure out some way of describing the state machines |
| 05:37.27 | beewoolie-afk | what state machine? |
| 05:37.27 | [g2] | maybe a table format is best |
| 05:37.37 | beewoolie-afk | The TAP state machine is well defined by the spec. |
| 05:38.13 | [g2] | Well there's a state machine processing commands |
| 05:38.18 | beewoolie-afk | The FPGA only needs to keep track of a coupe of things. |
| 05:38.18 | [g2] | in the FPGA right ? |
| 05:38.41 | beewoolie-afk | It needs to track the state of the TAP. This is done with a table that I wrote, but it is easy to derive. |
| 05:38.58 | beewoolie-afk | And, it needs to keep the last values of the TAP pins that it drove. |
| 05:39.18 | beewoolie-afk | This way, all it does is check for TCK rising and then it latches the new state. |
| 05:39.29 | beewoolie-afk | The only place it uses this is in the SEEK_STATE opcode. |
| 05:39.38 | beewoolie-afk | The rest is basically stateless. |
| 05:39.56 | beewoolie-afk | Even BULKIO doesn't need to do more than loop on the bits. |
| 05:40.08 | [g2] | well the FPGA has states regarding the line transitions |
| 05:40.36 | [g2] | where clock is, where the clocking domains are, etc... |
| 05:40.50 | [g2] | not to mention there's host communications |
| 05:41.54 | [g2] | Or maybe I'm thinking implementation details too quickly |
| 05:41.55 | beewoolie-afk | Not really. |
| 05:42.17 | beewoolie-afk | Well, perhaps a little, but I think that can be handled with simple logic. |
| 05:42.29 | beewoolie-afk | Again, the protocol doesn't need those things. |
| 05:42.48 | beewoolie-afk | Each of the instructions, except for SEEK_STATE, can be executed the same way regardless of the system state. |
| 05:42.50 | [g2] | you are correct that the protocol is nearly stateless |
| 05:43.06 | beewoolie-afk | Even the RESETS are handled that way. |
| 05:43.13 | beewoolie-afk | That's by design. |
| 05:43.25 | [g2] | you're talking like you are executing on a processor :) |
| 05:43.30 | beewoolie-afk | We may introduce some state when we start with comparesand loops. |
| 05:43.31 | [g2] | it's just lots of logic |
| 05:44.00 | beewoolie-afk | I confident of this because I've implemented most of this in code already. |
| 05:44.13 | ka6sox | the VHDL can do that. |
| 05:44.31 | ka6sox | as long as the commands are defined (as they are) |
| 05:44.49 | beewoolie-afk | 00000070: 00510051 00520052 00590059 00010001 Q.Q.R.R. Y.Y..... |
| 05:44.55 | beewoolie-afk | Do you know what that means???? |
| 05:45.28 | ka6sox | brb..phone |
| 05:45.29 | beewoolie-afk | I'm close to getting the flashing code working. |
| 05:46.17 | [g2] | with the parallel port ? |
| 05:46.20 | [g2] | or the usb ? |
| 05:48.32 | beewoolie-afk | parallel port. |
| 05:48.40 | beewoolie-afk | But the USB is just a skip away. |
| 05:48.53 | beewoolie-afk | I realized today that I *do* already have all o fthe parts I need. |
| 05:49.03 | beewoolie-afk | dollface is going to help me do the soldering. |
| 05:49.20 | beewoolie-afk | She's very excited about it. |
| 05:51.56 | [g2] | wow |
| 05:52.17 | beewoolie-afk | lol |
| 05:52.53 | [g2] | my wife call that "welding" :) |
| 05:53.13 | [g2] | maybe it's just the way _I_ solder :) |
| 05:53.18 | beewoolie-afk | lol |
| 05:54.29 | beewoolie-afk | I'd have finished this code already if it weren't for this ATP stuff. |
| 05:58.53 | beewoolie-afk | Nite folks. |
| 05:59.03 | *** part/#openjtag beewoolie-afk (n=beewooli@206.124.142.26) |
| 05:59.51 | *** join/#openjtag nslu2-log_ (n=nslu2-lo@nslu2-linux/dyoung) |
| 06:01.52 | [g2] | nite sweet dreams |
| 06:03.05 | ka6sox | nite |
| 06:03.43 | *** join/#openjtag nslu2-log_ (n=nslu2-lo@nslu2-linux/dyoung) |
| 12:37.36 | key2 | lennert: there? |
| 12:46.15 | [g2] | lennert congrats ! |
| 12:46.59 | lennert | thanks! |
| 12:55.43 | *** join/#openjtag prpplague (n=billybob@72.22.141.137) |
| 13:13.42 | key2 | lennert: u've succeded to run a linux on a PPC of a Xilinx ? |
| 13:19.25 | lennert | well |
| 13:19.31 | lennert | i wouldn't use a 'divisor' for the tck rate |
| 13:19.49 | key2 | lennert: why ? |
| 13:19.54 | lennert | divisor implies a base speed (divisor=1) and just isn't too fine-grained enough |
| 13:20.09 | lennert | i.e. what if your base rate is 50mhz and you want to run at 40? you have to run at 25.. |
| 13:20.36 | lennert | if the device implements it with a clock divider, it should be free to choose the next lower speed, i think |
| 13:20.58 | lennert | i like that it's a byte protocol |
| 13:21.06 | lennert | what i didn't find is what happens when there is a crc error |
| 13:21.08 | key2 | why would you reduce the speed of your PPC to run a linux ? |
| 13:21.24 | lennert | you throw the frame away.. and the other end will wait indefinitely? |
| 13:21.38 | vmaster | loool |
| 13:21.43 | key2 | is it a robot ? |
| 13:21.47 | key2 | i'm talkin to a robot ? |
| 13:22.06 | key2 | sounds like the answer have nothing to see with the questions :) |
| 13:22.18 | key2 | vmaster ? :) |
| 13:22.30 | vmaster | key2: this is just a guest, but i remember him introducing you to his ignore list |
| 13:22.34 | vmaster | -t |
| 13:22.37 | vmaster | +s |
| 13:22.40 | lennert | as to SEEK_STATE.. there can be many ways of arriving in a certain state |
| 13:23.10 | key2 | vmaster: really ? |
| 13:23.13 | key2 | hahah |
| 13:23.30 | key2 | talk to him see if he answers :) |
| 13:24.11 | lennert | vmaster: what's funny? |
| 13:24.51 | vmaster | lennert: key2 thought you were answering his question |
| 13:25.13 | vmaster | [14:14] < key2> lennert: u've succeded to run a linux on a PPC of a Xilinx ? |
| 13:25.13 | vmaster | [14:19] < lennert> well |
| 13:25.13 | vmaster | [14:19] < lennert> i wouldn't use a 'divisor' for the tck rate |
| 13:25.14 | vmaster | [14:20] < key2> lennert: why ? |
| 13:25.53 | vmaster | but obviously he's still on your ignore list, so you didn't see this |
| 13:27.12 | *** join/#openjtag cpage (n=cpage@netblock-68-183-40-90.dslextreme.com) |
| 13:28.25 | lennert | right, indeed |
| 13:32.58 | key2 | vmaster: past that |
| 13:33.10 | key2 | lennert: could u remove me from the ignore list ? |
| 13:51.49 | *** join/#openjtag GyrosGeier (n=richter@p5499682A.dip.t-dialin.net) |
| 15:14.13 | *** join/#openjtag bullet (n=bullet@99.50.62.81.cust.bluewin.ch) |
| 16:09.24 | *** join/#openjtag velinp (n=velinp@c88-163.cable.netissat.bg) |
| 16:17.08 | ka6sox | lennert, morning |
| 16:19.30 | lennert | ka6sox: morning |
| 16:24.19 | velinp | lennert, ka6sox: morning, guys; (evening would be more appropriate here in BG :) |
| 16:28.51 | velinp | managed to cross-compile apex-1.3.13; played a bit with make menuconfig; Marc Singer said yeah; but my problem is I still can't test it - hdw not ready :(; |
| 16:39.53 | *** join/#openjtag ByronT-Away (i=byron-po@nslu2-linux/ByronT) |
| 16:45.27 | ka6sox | velinp, w00t |
| 16:45.47 | ka6sox | Lennert: does the V2p have Rocket I/O? |
| 16:46.41 | velinp | ka6sox, now looking thru apex to see what CONFIG_SDRAM_BANK1 really does; |
| 16:46.49 | AchiestDragon | most of the v2p range have |
| 16:47.38 | AchiestDragon | some of the package options dont have any rocket io ports |
| 16:48.24 | AchiestDragon | morning btw :) |
| 16:50.23 | ka6sox | velinp, cool |
| 16:50.26 | ka6sox | AchiestDragon, thanks. |
| 16:50.44 | ka6sox | time for me to "start" my day....cya laters online here. |
| 18:01.42 | *** join/#openjtag beewoolie-afk (n=beewooli@206.124.142.26) |
| 18:12.30 | lennert | ka6sox: yeah, i think it does have (rocket i/o) |
| 18:12.45 | lennert | velinp: good evening |
| 18:14.27 | lennert | velinp: or as they say, Добър вечер |
| 18:14.42 | velinp | lennert: I can't read that |
| 18:15.04 | velinp | lennert: please, transcribe |
| 18:15.10 | lennert | dober vecher |
| 18:15.21 | lennert | more like dob"r |
| 18:15.37 | velinp | lennert: dobyr vecher; glad to see/hear you |
| 18:16.47 | velinp | lennert: how are you doing? |
| 18:16.53 | lennert | problem with transliterating cyrillic is that everyone does it differently.. |
| 18:17.13 | lennert | dobpe sym? s"m? |
| 18:17.17 | lennert | dobre |
| 18:17.28 | lennert | some people even write sha as 'w' |
| 18:17.34 | lennert | because the w looks like a sha |
| 18:17.57 | velinp | yes, but cp-1251 seems to be the std (for BG); how does it happen your bulgarian is quite good |
| 18:18.16 | lennert | right, cp1251. that's also what allofmp3.com uses for russian tracks. |
| 18:18.30 | lennert | it's really annoying if your mp3 player insists on utf8 |
| 18:18.51 | lennert | well, i know a bit of russian, and like 10 words of bulgarian |
| 18:19.05 | velinp | been in bg? |
| 18:19.15 | lennert | nope |
| 18:19.44 | velinp | intend to? nice ski/sea here; not that i do either much ... |
| 18:19.50 | lennert | hehe |
| 18:19.54 | lennert | i live 5km from the sea where i am |
| 18:20.01 | lennert | (it's a bit cold here though) |
| 18:20.11 | lennert | i came back on monday from lithuania and there was lots of snow there |
| 18:20.37 | velinp | baltic; i've dipped a finger in it; you parents in law live there; you in nl? |
| 18:20.51 | lennert | yes |
| 18:21.00 | lennert | note that lithuanian and latvian are very different from estonian |
| 18:21.20 | velinp | it's good to have you mother-in-law in a different country :) |
| 18:21.24 | lennert | haha |
| 18:21.28 | lennert | yes, sometimes :) |
| 18:21.51 | lennert | well, my mother in law is a very kind woman, it's mostly the culture differences that i can't get used to |
| 18:21.57 | lennert | like, them offering me food every 5 minutes |
| 18:22.03 | *** join/#openjtag Tiersten (n=tman@nslu2-linux/Tiersten) |
| 18:22.08 | velinp | instead of a drink? :) |
| 18:22.16 | lennert | i take the drinks but not the food :) |
| 18:22.27 | lennert | i mean, it's like we speak different languages |
| 18:22.47 | lennert | "do you want to eat X?" "no, thank you, i'm not hungry." "do you want to eat Y?" didn't i just say i wasn't _hungry_? |
| 18:23.08 | velinp | there is a nice bulgarian word - meze; it's same as the russian zakusivat |
| 18:23.40 | lennert | my dictionary doesn't have that word.. |
| 18:24.18 | velinp | i remember m. sholohov; not before the third glass ... |
| 18:25.07 | lennert | which of them?? :) |
| 18:26.10 | velinp | it was about a nazy camp; the guy was 'invited'; he had to drink in order to take some food to his fellow prisoner's; not sure by m.s. |
| 18:27.02 | lennert | i got quite a lot of practise in the last few weeks |
| 18:28.18 | velinp | me too; but i do it 'moderately'; i mean i drink with my wife :) |
| 18:28.39 | *** join/#openjtag [g2-lap] (n=g2@cpe-066-057-008-035.nc.res.rr.com) |
| 18:28.42 | lennert | and when she drops down drunk, you stop too? :) |
| 18:29.10 | [g2-lap] | lennert I thought you'd get a kick out of this http://www.picocomputing.com/products/cards.php |
| 18:30.19 | lennert | [g2-lap]: looks like it'd be expensive |
| 18:31.13 | [g2] | it is, but that doesn't make it totally uncool (around $2.8K) |
| 18:31.58 | lennert | well, yeah |
| 18:32.15 | lennert | $2800, jeeez |
| 18:32.35 | [g2] | but it is fun to daydream about plugging one into the Loft's CF slot |
| 18:33.10 | lennert | sure |
| 18:33.20 | lennert | i like daydreaming about having $RANDOM_EXPENSIVE_THING |
| 18:33.35 | [g2] | yeah it's fun for a little while |
| 18:37.38 | velinp | lennert: thanks for the chat; how do you get this slant font (like [g2] goes back to work)? |
| 18:37.54 | lennert | velinp: use /me ? |
| 18:37.57 | lennert | /me blah |
| 18:38.32 | velinp | lennert: ok, thanks; |
| 18:40.08 | wookey__ | (OT = I went to bulgaria skiing a few years ago - the skiing was crummy (not much snow) but the partying was excellent. A fine country. |
| 18:40.27 | wookey__ | particularly liked the sweet red wines |
| 18:41.04 | velinp | wookey__: I like it too; except some things :) |
| 18:43.44 | prpplague | wookey__: hehe i want to visit belgium sometime to go on a tour of the abbeys |
| 18:44.48 | prpplague | lennert: the lnode80 is only slightly larger than a cf card |
| 18:45.00 | prpplague | lennert: basically its the same size as a 802.11b cf card |
| 18:46.07 | lennert | i particularly like the liqour in lithuania |
| 18:46.13 | lennert | its most redeeming quality being that it's cheap |
| 18:46.21 | lennert | and available everywhere at any time of day |
| 18:46.22 | prpplague | hehe |
| 18:46.40 | prpplague | lennert: hehe, kind of like rum here in barbados |
| 18:46.46 | ka6sox | hic |
| 18:47.11 | ka6sox | and Taquila here |
| 18:49.54 | lennert | in holland mostly just wine |
| 18:50.21 | lennert | i fell in love with vodka the first time i went to .lt, about 3.5 yrs ago |
| 18:50.46 | ka6sox | uh oh...at least you haven't hit the Schnapps. |
| 18:50.59 | lennert | i can't drink schnapps |
| 18:51.19 | ka6sox | and Gin just is revolting |
| 18:51.32 | ka6sox | (mother liked Gin) |
| 18:52.07 | lennert | agree |
| 18:52.26 | lennert | the less contaminated the alcohol is, the better |
| 18:53.55 | ka6sox | okay enough of drinking before local noon...back to work! |
| 18:53.58 | ka6sox | bye for now |
| 18:56.30 | beewoolie-afk | Hard to believe that PICO card is $2.8k |
| 18:57.47 | lennert | ah, beewoolie-afk |
| 18:57.51 | lennert | i wrote some comments on your proposal |
| 18:57.55 | lennert | but i think you weren't there when i did |
| 18:58.25 | beewoolie-afk | lennert: share? |
| 18:58.37 | lennert | < lennert> i wouldn't use a 'divisor' for the tck rate |
| 18:58.39 | beewoolie-afk | BTW, I updated part of it since the DELAY opcode wasn't useful. |
| 18:58.43 | lennert | < lennert> divisor implies a base speed (divisor=1) and just isn't too fine-grained enough |
| 18:58.50 | lennert | < lennert> i.e. what if your base rate is 50mhz and you want to run at 40? you have to run at 25.. |
| 18:58.50 | beewoolie-afk | Certainly. |
| 18:58.58 | lennert | < lennert> if the device implements it with a clock divider, it should be free to choose the next lower speed, i think |
| 18:59.01 | lennert | < lennert> i like that it's a byte protocol |
| 18:59.04 | lennert | < lennert> what i didn't find is what happens when there is a crc error |
| 18:59.08 | beewoolie-afk | We could use an integer/fraction encoding. |
| 18:59.12 | lennert | < lennert> you throw the frame away.. and the other end will wait indefinitely? |
| 18:59.31 | beewoolie-afk | CRC errors should return a result indicating that the transaction failed. |
| 18:59.34 | lennert | beewoolie-afk: i would just specify the rate in mhz or something |
| 18:59.46 | beewoolie-afk | Problem is that we may not know the ID if it's garbled. |
| 18:59.48 | lennert | beewoolie-afk: but what if the result packet has a crc error/ |
| 19:00.02 | beewoolie-afk | Yeah, that's always a problem. |
| 19:00.19 | beewoolie-afk | At least, this way we can detect the error. |
| 19:00.24 | vmaster | is a crc error really a possible scenario? |
| 19:00.35 | beewoolie-afk | Well, it's good policy and it's cheap. |
| 19:00.47 | lennert | vmaster: if you run it over serial, sure.. |
| 19:01.05 | beewoolie-afk | Bits in memory *can* rot. |
| 19:01.26 | lennert | but crc/fcs protects against transmission errors.. |
| 19:02.08 | vmaster | okay, but this protocol is to be used on a short point-to-point connection, between the uC and the fpga, right? |
| 19:02.35 | beewoolie-afk | Ever read this: http://research.microsoft.com/~lampson/33-Hints/WebPage.html |
| 19:03.14 | beewoolie-afk | Read 4. fault tolerance |
| 19:05.54 | beewoolie-afk | Error detection and recovery should be end-to-end. |
| 19:06.19 | beewoolie-afk | Depending on intermediate checks doesn't give the system reliability. |
| 19:07.11 | beewoolie-afk | lennert: The problem with setting the clock frequency as a frequency is that it may not be achievable. |
| 19:07.21 | lennert | beewoolie-afk: so you choose a lower one |
| 19:07.26 | vmaster | yeah, hence my question if it's really a possible scenario that data might get corrupted in our setup |
| 19:07.34 | beewoolie-afk | Yeah, but the user may nowbe able to kow it. |
| 19:07.54 | lennert | beewoolie-afk: the device might not support divisor=1 either |
| 19:08.00 | beewoolie-afk | vmaster: that's a short sighted point of view. |
| 19:08.11 | beewoolie-afk | The CRC is cheap and it guarantees correctness. |
| 19:08.28 | beewoolie-afk | Without out it, there is no way to know if the intervening layers did the right thing. |
| 19:10.42 | beewoolie-afk | And, the cost of including it is negligible. |
| 19:11.07 | beewoolie-afk | lennert: of course it supports divisor 1. How coult it not? |
| 19:12.02 | beewoolie-afk | lennert: I don't really care how to encode the frequency, but it need to be such that most or all of the encodable values are valid. |
| 19:12.51 | beewoolie-afk | It is sloppy to encode 80% of values that don't do what the user requests. |
| 19:54.25 | beewoolie-afk | lennert: Do you understand how we have to handle missing transactions? |
| 20:03.57 | ka6sox-office | hi |
| 20:04.02 | ka6sox-office | in office now. |
| 20:04.29 | ka6sox-office | Lennert: ping? |
| 20:06.08 | ka6sox-office | you can use either dividers OR DPLL's in the FPGA's to provide clocks. |
| 20:06.43 | beewoolie-afk | I can think of two good questions. |
| 20:06.58 | beewoolie-afk | 1) what values for tclk frequency are likely to be needed? |
| 20:07.12 | beewoolie-afk | 2) how accurately will the user expect to set these values? |
| 20:08.22 | vmaster | ah, RTCK would be useful, too |
| 20:08.22 | beewoolie-afk | The BSDL files I've seen have a maximum clock value, usually in the for Vx10^E where V is a small integer and E is usually 6. |
| 20:08.30 | beewoolie-afk | What is RTCK? |
| 20:08.39 | vmaster | the returned TCK |
| 20:08.47 | vmaster | synthesizable arm cores use it |
| 20:09.05 | beewoolie-afk | Can you explain a little more? |
| 20:09.30 | beewoolie-afk | Is this the variable clocking that some JTAG emulators refer to? |
| 20:09.45 | vmaster | the TCK enters the ARM and gets syncronized with the processor clock, and this is output on RTCK |
| 20:10.15 | vmaster | they're using adaptive clock for two different things - one is rtck, and one is driving the tck as high as long as the device can follow |
| 20:10.51 | beewoolie-afk | ah, so they drive tck faster until rtck doesn't keep up? |
| 20:11.05 | ka6sox-office | I haven't seen a lot of devices with RTCLK |
| 20:11.19 | beewoolie-afk | ka6sox-office: Have you seen any? |
| 20:11.45 | ka6sox-office | on SOME MIPS and a few of the Freescale procs. |
| 20:11.58 | ka6sox-office | I don't know ARM well enough |
| 20:12.19 | beewoolie-afk | would this be ejtag? |
| 20:12.27 | vmaster | well, i only know it from arm |
| 20:12.54 | beewoolie-afk | vmaster: any example cores we can look at? |
| 20:13.04 | ka6sox-office | I don't know about the Xscale |
| 20:13.06 | vmaster | ARM7TDMI-S, ARM926EJ-S are the most common |
| 20:13.26 | vmaster | mhh, xscale is a hard core, so it doesn't need the rtck |
| 20:13.28 | beewoolie-afk | Is the -S the variant with this feature? |
| 20:13.40 | vmaster | all -S variants require rtck |
| 20:13.50 | vmaster | you can go without, but that limits you to 1/6th pclk |
| 20:13.56 | vmaster | http://www.arm.com/support/faqip/3732.html |
| 20:15.39 | beewoolie-afk | Well, 1/6 of 200MHz is 33MHz. Far better than any JTAG emulator that I've used so far. |
| 20:16.36 | ka6sox-office | we might have devices that can't go that fast either |
| 20:16.43 | ka6sox-office | soooo...we may need to divide them down. |
| 20:17.34 | ka6sox-office | we would use the divided clock for creating the TCLK that would drive our logic. |
| 20:18.03 | ka6sox-office | we might have to go as low as 4mhz |
| 20:18.12 | ka6sox-office | (wouldn't be for something like the Xscale |
| 20:18.53 | ka6sox-office | the ARM9 core on the TS7200 is capable of this clock change? |
| 20:19.14 | beewoolie-afk | ka6sox-office: BTW, I added that link, but it needs approval. I suppose this is goatse.cx protection? |
| 20:20.31 | beewoolie-afk | I can see that RTCK is an important feature to support. The only way to test that it works, though, is to have a core that requires it. |
| 20:20.41 | AchiestDragon | if there is a device data base for the chip types , then the max tclock should be in the data , so it could be made to set up auto for that |
| 20:21.12 | vmaster | get a 30$ lpc board - they are all arm7tdmi-s |
| 20:21.18 | AchiestDragon | you can software set the clock on the ts7200 |
| 20:21.20 | ka6sox-office | beewoolie-afk, yes thats right |
| 20:21.27 | ka6sox-office | I need Tiersten to approve it :) |
| 20:23.02 | vmaster | and the 1/6th seems to be best-case - i can't reliably debug an lpc2294 running at 60mhz with the ft2232c at 6mhz |
| 20:23.22 | beewoolie-afk | I found this interesting page in the FAQ. |
| 20:23.28 | ka6sox-office | vmaster, how many bits in the bsdl? |
| 20:23.39 | beewoolie-afk | ARM uses a file in the format that I was originally thinking of for test. |
| 20:23.44 | beewoolie-afk | http://www.arm.com/support/faqip/3845.html |
| 20:24.11 | beewoolie-afk | I'm not now sure this is the right way for us to handle test, but it is easy. |
| 20:24.20 | vmaster | ka6sox-office: the lpcs have no boundary-scan chain, the debug register is 33 bits long |
| 20:25.30 | vmaster | beewoolie-afk: did you read the svf spec? |
| 20:26.00 | beewoolie-afk | I read it over, but I didn't comprehend what they were doing. I may have missed part of the spec. |
| 20:26.38 | ka6sox-office | vmaster, ah |
| 20:27.01 | vmaster | well, all they really do is specify states in which to end a scan, and then the IR and DR scans, the value to be scanned in, and the expected result |
| 20:27.15 | lennert | beewoolie-afk: if you have a high-speed and a low-speed jtag device, then one of them won't be able to support divisor=1, or the high speed one can't run faster than low speed |
| 20:27.43 | beewoolie-afk | lennert: That's true no matter how the frequency is encoded. |
| 20:27.52 | vmaster | lennert: with this protocol you're likely to have only device |
| 20:27.58 | ka6sox-office | ype |
| 20:28.04 | lennert | 20:11 < beewoolie-afk> lennert: of course it supports divisor 1. How coult it not? |
| 20:28.11 | lennert | just responding to your statement |
| 20:28.15 | ka6sox-office | so a slow CPLD would slow down a fast proc |
| 20:28.21 | beewoolie-afk | Yeah, but that's the same as saying FREQ=40MHz. |
| 20:28.44 | lennert | sigh |
| 20:28.59 | beewoolie-afk | feeling frustrated? |
| 20:29.25 | beewoolie-afk | I agree that some parts won't go as fast as others. |
| 20:29.44 | beewoolie-afk | The questions is how do we describe the TCLK rate. |
| 20:29.48 | ka6sox-office | we will have to have a method for determining what speed we can sustain. |
| 20:30.14 | ka6sox-office | beewoolie-afk, its a register value in the fpga that is written by the proc. |
| 20:30.17 | beewoolie-afk | We can use a slow rate to read the IDCODES of the core on the chain. |
| 20:30.32 | beewoolie-afk | Once we ID the cores, we know what rates they can sustain. |
| 20:30.51 | beewoolie-afk | Then it's a matter of setting the rate. |
| 20:31.12 | ka6sox-office | we can pass that information to the TS and then have it set a register. |
| 20:31.16 | beewoolie-afk | lennert suggested that we specify this in MHz. I'm OK with that as long as we can really set the rate to the value requested. |
| 20:31.37 | ka6sox-office | I think that we will have divisors that can be defined. |
| 20:31.46 | ka6sox-office | (and thus we can set it by mhz) |
| 20:32.09 | beewoolie-afk | ka6sox-office: this was about the use of a divisor in the protocol. We just need to make the protocol describe something that the FPGA can use. |
| 20:32.28 | beewoolie-afk | It could be a predefined table of values. |
| 20:32.50 | ka6sox-office | if we define the reference oscillator frequency we can do that. |
| 20:33.27 | beewoolie-afk | ka6sox-office: remember, this is about making the protocol make sense. We don't want to hard-code the protocol so that it only works with this version of the FPGA. |
| 20:33.38 | beewoolie-afk | The best method is something that is flexible and clear. |
| 20:33.53 | beewoolie-afk | Divisors aren't so good in the protocol because they're dependent on FPGA implementation. |
| 20:33.59 | ka6sox-office | the VHDL code is mostly portable with *some* specific things related to the hardware. |
| 20:34.29 | beewoolie-afk | ka6sox-office: It isn't about that. We may implement the protocol with something other than hardware. |
| 20:34.47 | vmaster | what could that be? |
| 20:34.47 | beewoolie-afk | e.g. SW implementation to drive parallel port dongle. |
| 20:35.26 | ka6sox-office | the parallel port dongle is just a buffer...doesn't have *any* logic |
| 20:35.38 | beewoolie-afk | and so... |
| 20:35.58 | beewoolie-afk | We still need a way to verify testing procedures. |
| 20:36.13 | ka6sox-office | we have logic and we want to set the frequency..that is done by setting a register that controls a divisor. |
| 20:36.19 | beewoolie-afk | Right. |
| 20:36.26 | *** join/#openjtag ByronT (i=byron-po@nslu2-linux/ByronT) |
| 20:36.59 | beewoolie-afk | Look at it this way. |
| 20:37.01 | *** join/#openjtag prpplague (n=billybob@72.22.141.137) |
| 20:37.03 | ka6sox-office | the way that is set depends on what devices you have in your toolkit (for each FPGA type) |
| 20:37.11 | beewoolie-afk | Let's say we implement a second version that is *way* faster than the first. |
| 20:37.29 | beewoolie-afk | The first one can only go 4 MHz because we borked it. It's still useful, tho. |
| 20:37.50 | beewoolie-afk | it would be helpful if there were a way for the controlling software to gracefully handle these cases. |
| 20:38.11 | beewoolie-afk | We want to be able to cope with changes in the interpreting hardware. |
| 20:38.42 | ka6sox-office | the clock freq and divisors are the determinant elements. |
| 20:38.58 | ka6sox-office | thats a function of hardware. |
| 20:39.03 | beewoolie-afk | so, the second implementation can go to 11 (i.e. 400MHz). there shouldn't be anything precluding a testing stream that works OK with this. |
| 20:39.21 | beewoolie-afk | then perhaps we need a way to query the hardware to know how it handles clocking? |
| 20:39.32 | beewoolie-afk | Ask the FPGA what it's base frequency is? |
| 20:39.57 | ka6sox-office | that would have to be a hard coded thing or we can just say "the clock freq is 40mhz" period. |
| 20:40.18 | beewoolie-afk | ka6sox-office: But if we plug in another device that can go faster, how is the SW going to know this? |
| 20:40.21 | ka6sox-office | we can then run the FPGA at 240mhz interanl. |
| 20:40.32 | ka6sox-office | faster than 40mhz? |
| 20:41.13 | ka6sox-office | above 100mhz wire length becomes an issue. |
| 20:41.14 | beewoolie-afk | OK. It is fine to say that the base frequency is 40MHz. Period. |
| 20:41.27 | ka6sox-office | from that we can divide down |
| 20:41.27 | beewoolie-afk | do we then use a divider? |
| 20:41.33 | ka6sox-office | yes |
| 20:41.37 | beewoolie-afk | :-) Back to the beginning! |
| 20:41.44 | ka6sox-office | yep |
| 20:41.47 | vmaster | can we divide by fractions? |
| 20:41.55 | beewoolie-afk | Should the divider be INT/FRAC? |
| 20:41.59 | ka6sox-office | sometimes the clock dividers can do that. |
| 20:42.19 | ka6sox-office | I would have to look at the S3 (but it may) |
| 20:45.26 | ka6sox-office | phone |
| 21:05.10 | ka6sox-office | off phone |
| 22:44.37 | AchiestDragon | guess that it could be easy to do a auto max speed detect ,, |
| 22:44.55 | ka6sox-office | ya |
| 22:45.27 | ka6sox-office | the best thing would be to read the ID_CODES and then have the Proc set the max value for the chain. |
| 22:45.56 | AchiestDragon | run a jtag read , at a low speed , run it again comparing the results at a faster speed ,,, and again , untill it starts showing diferences , so then drop it back one |
| 22:46.39 | beewoolie-afk | yeah, that's what I wrote. |
| 22:46.51 | ka6sox-office | which one? |
| 22:46.52 | ka6sox-office | :) |
| 22:47.34 | ka6sox-office | they should both work. |
| 22:47.49 | beewoolie-afk | And that's fine as far as it goes, but a) we have to have accurate BSDL files, and b) the user still needs to be able to change the speed. |
| 22:48.05 | AchiestDragon | as far as the speed setting to the program it would be a register in the fpga , could be read write so that the program will be able to know the speed |
| 22:48.11 | beewoolie-afk | As far as I can tell from my very small sample, BSDL files are notorously error prone. |
| 22:48.29 | ka6sox-office | beewoolie-afk, I suspect you are correct. |
| 22:48.45 | beewoolie-afk | AchiestDragon: but remember that the issue is that we need to put something into the protocol to allow that register to be written. |
| 22:48.48 | ka6sox-office | *if* you are using a file with the app from the vendor (proprietary) it works. |
| 22:49.09 | beewoolie-afk | Uh, huh. So, you mean to steal those files? |
| 22:49.23 | beewoolie-afk | I've been musing about this for some time. |
| 22:49.28 | ka6sox-office | sometimes they are available and sometimes not |
| 22:49.30 | beewoolie-afk | BSDL files will always be a problem. |
| 22:49.38 | ka6sox-office | yep |
| 22:49.44 | AchiestDragon | plus if the spec for a chip says it can do 40mhz it may only work at 30mhz in the circut the device under test is fitted to |
| 22:49.48 | beewoolie-afk | My plan has been to augment BSDL files with data that helps us detect the specifics. |
| 22:49.57 | ka6sox-office | good plan |
| 22:50.05 | ka6sox-office | that way we can flesh them out. |
| 22:50.09 | ka6sox-office | brb |
| 22:50.12 | beewoolie-afk | for example, there may be several cores with the same ID, but the target has variations. |
| 22:50.39 | beewoolie-afk | So, we have detect clauses in the extra meta-data that helps us determine exactly which target we're talking to. |
| 22:50.50 | beewoolie-afk | for example, we can read CP15 registers on ARM cores. |
| 22:51.24 | AchiestDragon | and if a board has a chain of them there may be diferent devices with diferent max clocking rates , plus delays that may accur with chaining them |
| 22:51.29 | beewoolie-afk | And, SoCs have more registers we can poll to make sure we're talking to the specific vendor's chip. |
| 22:51.38 | beewoolie-afk | Of course. |
| 22:51.53 | beewoolie-afk | But we have to scan out all of the IDCODES and probe each one to find out. |
| 23:04.30 | ka6sox-office | sounds like fun |
| 23:06.21 | beewoolie-afk | vmaster: ping? |
| 23:42.47 | AchiestDragon | ka6sox-office: ping |
| 23:43.52 | beewoolie-afk | 8-) |
| 23:44.40 | AchiestDragon | ok the ram tracks fin on 2 layers , the pc104 with the link options for ts7200 or std pc104 seems to work on 2 layers ( still got some signals to test route there ) |
| 23:45.11 | AchiestDragon | and got 18 i/o lines for the jtag / analizer interface to play with |
| 23:45.22 | AchiestDragon | so looking good |
| 23:45.30 | ka6sox-office | w00t |
| 23:47.19 | ka6sox-office | 18 lines is good because we can have triggers :) |
| 23:47.49 | AchiestDragon | problem or portential problem i spotted , on the jtag connector as used the arm20pin format is not the same as the ieee std one http://www.interfacebus.com/Design_Connector_JTAG_Bus.html |
| 23:48.09 | AchiestDragon | there diferent pinouts |
| 23:49.43 | AchiestDragon | well going to need 1 bit for selecting the analizer buffers from that 18 |
| 23:50.36 | ka6sox-office | the 20 pin ARM one has all the functions. |
| 23:50.54 | AchiestDragon | so does the ieee one |
| 23:51.10 | ka6sox-office | AchiestDragon, I understand thats why I said we have a bit for a trigger :) |
| 23:51.22 | AchiestDragon | k |
| 23:52.00 | ka6sox-office | (assuming we have 16 bits we care to capture) |
| 23:52.28 | AchiestDragon | have been thinking about it also as i would like it to support the 10 pin jtag version that is used oftern also |
| 23:53.41 | AchiestDragon | yes , you should be able to triger on a state of any of the bits or a specific state of 2 or more bits |
| 23:54.47 | AchiestDragon | think the extra 17th bit would be better use if it was a clock input so it could sync to the unit its analizing |
| 23:55.29 | ka6sox-office | ep1220's board has adapters to make this all happen. |
| 23:55.39 | ka6sox-office | 5 or 6 different formats. |
| 23:56.20 | ka6sox-office | I think that all the common ones are covered. |
| 23:56.33 | AchiestDragon | k, have been thinking about useing the extra 100mm*60mm spare pcb space for some adaptors |
| 23:57.53 | ka6sox-office | next time you see ep1220 ask him about it...he may not have placed his order and we could put them on your spare space. |
| 23:58.11 | ka6sox-office | he shows up fairly regularly and is in .at. |