| 00:13.34 | beewoolie-afk | vmaster: I don't have the benefit of chat logs, but I'm quite certain you've made the performance of the 2232, especially with respect to aggregate throughput. |
| 00:34.03 | vmaster | mhh, there's always ibot.rikers.org for your log needs |
| 00:34.36 | ka6sox-office | I keep forgetting about that. |
| 00:36.39 | vmaster | regarding the ft2232: when doing blind writes to the dcc at 6mhz tck i achieve about 120kByte/s - one write to the dcc register requires 51 tck cycles (7x tms, 37x tdi, 7x tms), so this is about 1.5 MHz |
| 00:37.56 | *** join/#openjtag key2 (n=key2@gob75-2-81-56-64-17.fbx.proxad.net) |
| 00:43.37 | vmaster | okay, n8 everyone, i'm off |
| 01:02.47 | *** join/#openjtag Tiersten_ (n=tman@nslu2-linux/Tiersten) |
| 01:17.43 | *** join/#openjtag Tiersten (n=tman@nslu2-linux/Tiersten) |
| 01:21.49 | *** join/#openjtag vmaster_ (i=vmaster@p549B78BD.dip.t-dialin.net) |
| 01:30.01 | beewoolie-afk | nite vmaster |
| 01:30.17 | beewoolie-afk | The speed he is reporting, 120KB/s is really quite good. |
| 01:31.08 | beewoolie-afk | What I think may be missing is the aggretate throughput when polling is required. |
| 01:32.11 | beewoolie-afk | fortunately, ARM figured this out at some point and improved the efficiency of polling. |
| 03:22.00 | *** join/#openjtag ka6sox (n=ka6sox@nslu2-linux/ka6sox) |
| 05:37.18 | *** join/#openjtag lennert (n=buytenh@alephnull.demon.nl) |
| 06:32.21 | beewoolie-afk | lennert: hi |
| 06:33.45 | ka6sox | nite beewoolie :) |
| 07:14.15 | lennert | beewoolie: hi and bye :) |
| 09:20.39 | *** join/#openjtag cpage__ (n=cpage@netblock-68-183-40-90.dslextreme.com) |
| 11:40.13 | *** join/#openjtag bullet (n=bullet@94.55.76.83.cust.bluewin.ch) |
| 12:21.15 | *** join/#openjtag prpplague (n=billybob@72.22.143.75) |
| 14:38.58 | *** join/#openjtag wookey__ (n=wookey@217.147.92.89) |
| 16:13.33 | *** join/#openjtag ByronT-Away (i=byron-po@nslu2-linux/ByronT) |
| 16:20.38 | *** join/#openjtag bullet (n=bullet@94.55.76.83.cust.bluewin.ch) |
| 16:49.09 | [g2] | any GUI / XML users in here ? |
| 17:46.27 | *** join/#openjtag velinp (n=velinp@c88-163.cable.netissat.bg) |
| 17:46.51 | velinp | ~seen ka6sox |
| 17:47.00 | purl | ka6sox <n=ka6sox@nslu2-linux/ka6sox> was last seen on IRC in channel #openjtag, 11h 13m 15s ago, saying: 'nite beewoolie :)'. |
| 17:47.52 | velinp | ~seen lennert |
| 17:47.53 | purl | lennert is currently on #openjtag (12h 10m 35s). Has said a total of 1 messages. Is idling for 10h 33m 38s, last said: 'beewoolie: hi and bye :)'. |
| 18:22.36 | lennert | i'm here |
| 18:54.27 | *** join/#openjtag Tiersten (n=tman@nslu2-linux/Tiersten) |
| 19:02.50 | *** join/#openjtag beewoolie-afk (n=beewooli@206.124.142.26) |
| 19:05.16 | beewoolie-afk | I built ep1220's board yesterday. |
| 19:05.29 | beewoolie-afk | Plugged it in, a.k.a 'the smoke test'. |
| 19:05.40 | beewoolie-afk | Bus 002 Device 004: ID 0403:6010 Future Technology Devices International, Ltd |
| 19:05.45 | beewoolie-afk | That's a good sign :-) |
| 19:07.29 | beewoolie-afk | Unfortunately, I forgot to buy headers for making ribbon cables. |
| 19:14.04 | *** join/#openjtag ByronT-Away (i=byron-po@nslu2-linux/ByronT) |
| 20:13.12 | *** part/#openjtag velinp (n=velinp@c88-163.cable.netissat.bg) |
| 20:56.48 | ka6sox-office | beewoolie-afk, w00hoo |
| 20:56.56 | beewoolie-afk | ka6sox-office: hey. |
| 20:57.08 | ka6sox-office | ya |
| 20:57.21 | beewoolie-afk | I've given up on te arm7 stuff for a while. It looks like there may be something fishy with the scan chains on the chips I have. |
| 20:57.32 | ka6sox-office | ah... |
| 20:57.41 | beewoolie-afk | I'm starting on the Xscale tomorrow. |
| 20:57.46 | ka6sox-office | AchiestDragon finished a pretty good S3 board design... |
| 20:57.58 | beewoolie-afk | I soldered up another slug with a jtag port. It gets easier every time. |
| 20:58.04 | ka6sox-office | a few are looking it over and then we will finalize it. |
| 20:58.08 | beewoolie-afk | w00t |
| 20:59.04 | vmaster | beewoolie-afk: what arm7s did you try? |
| 20:59.20 | beewoolie-afk | vmaster: I only have the Sharp parts, lh79520 lh79524. |
| 20:59.31 | beewoolie-afk | What's wierd is the scan chain 1 is only 33 bits long. |
| 20:59.36 | vmaster | yeah |
| 20:59.41 | beewoolie-afk | Their BSDL files are really borked, too. |
| 20:59.58 | vmaster | scan-chain 1 is supposed to be 33 bits on a arm7 |
| 21:00.01 | beewoolie-afk | It says that the instructions are only three bits long, but I can show that the register is 4 bits long. |
| 21:00.11 | beewoolie-afk | vmaster: ...hang on... |
| 21:00.31 | beewoolie-afk | Oh, right. That wasn't what didn't work. The CP15 register was shorted than expected. |
| 21:00.42 | beewoolie-afk | It is documented as being 32+4 bits, IIRC. |
| 21:00.48 | beewoolie-afk | It is reading out as 33 bits long. |
| 21:00.57 | vmaster | yeah, 33 bits is correct |
| 21:01.18 | vmaster | 32-bits data/instruction plus one bit that says if it's D or I |
| 21:01.19 | beewoolie-afk | Really? Not according to the 720T datasheet I was reading... |
| 21:01.29 | beewoolie-afk | For the CP15 register? |
| 21:01.31 | vmaster | yes |
| 21:02.15 | vmaster | regarding the instruction register size: there's usually a difference between bsdl mode and jtag mode |
| 21:02.15 | [g2] | when you guys get a minute I'd like to ask the brain trust a question |
| 21:02.31 | beewoolie-afk | Page 9-25 of the arm ddI0229a says 37 bits. |
| 21:03.02 | beewoolie-afk | Scan chain 15 has a read/write bit, instruction encoding bits 3:0 and data bust bits 31:0 |
| 21:03.21 | beewoolie-afk | vmaster: I'm looking at the JTAG scan chain documentation. |
| 21:03.37 | vmaster | the LH's are hard-core, not -S |
| 21:03.45 | beewoolie-afk | Yeah. |
| 21:03.58 | vmaster | ddi0229a is for the revision 4, using the -s core |
| 21:04.03 | beewoolie-afk | Hmm. |
| 21:04.26 | beewoolie-afk | How can you tell from the document? |
| 21:04.50 | vmaster | 1.1 About the ARM720t processor |
| 21:04.51 | beewoolie-afk | Oh, there it is. |
| 21:05.18 | beewoolie-afk | Unfortunately, Sharp doesn't point to the correct data sheet for their processor. |
| 21:05.26 | beewoolie-afk | I have to hunt until I find one that works. |
| 21:05.36 | vmaster | DDI0192A |
| 21:05.37 | beewoolie-afk | In fact, I'm guessing that they don't use the -S core. |
| 21:07.06 | beewoolie-afk | vmaster: thanks. |
| 21:08.13 | vmaster | np |
| 21:10.01 | beewoolie-afk | Hey look at that. All of the numbers match up now. :-o |
| 21:10.04 | vmaster | ah, the lh79520 tec datasheet explicitly mentions that it has a arm720t rev3 core |
| 21:10.41 | beewoolie-afk | Of course, I assumed that rev 4 was just rev 3 but better. |
| 21:10.52 | beewoolie-afk | It was working for the arm9 docs, for the most part. |
| 21:11.27 | beewoolie-afk | I think that once I found that their BSDL file was bogus, I assumed that everything was suspect. |
| 21:11.54 | beewoolie-afk | Even their published IDCODE is wrong. |
| 21:12.15 | vmaster | mhh, for the atmel parts, the bsdl doesn't describe the debug port at all - a pin decides on startup if the jtag port comes up as bsdl (bsdl file matches) or debug |
| 21:12.48 | ka6sox-office | vmaster: is this information in a wiki somewhere? |
| 21:12.52 | beewoolie-afk | [g2]: do you still have a Q? |
| 21:12.58 | [g2] | beewoolie-afk I do |
| 21:13.10 | [g2] | this is for all |
| 21:13.53 | [g2] | I'm curious if you guys are interested in some high-level form of the processor specific information |
| 21:14.15 | [g2] | example, I'd like to start playing with the IXP42x memory controller for the CS0/CS1 |
| 21:14.25 | [g2] | or pull all the CP 15 registers out |
| 21:15.30 | [g2] | getting a dump in a text/script friendly way (like the VxWorks ICE tool) or having a GUI makes sense to me for manipulating these values |
| 21:15.39 | [g2] | GPIO values are another example |
| 21:15.46 | [g2] | Thoughts ? |
| 21:16.36 | ka6sox-office | I'm just used to the way that the BDI2000 does it |
| 21:16.52 | ka6sox-office | its a telnet session. |
| 21:17.00 | ka6sox-office | so a gui would be nice. |
| 21:19.42 | [g2] | ka6sox-office you could use a sript within the telnet session |
| 21:20.06 | ka6sox-office | yep |
| 21:20.52 | beewoolie | Scripting telnet is dicey |
| 21:20.59 | beewoolie | The interfaces change and the code no longer works. |
| 21:21.19 | [g2] | I'd like to start pulling out all the CP15 info which includes Endianness and cache setup |
| 21:21.20 | beewoolie | Are you talking about a UBE solution, or something to make the final products work well? |
| 21:21.29 | beewoolie | For what purpose? |
| 21:21.52 | [g2] | beewoolie to me there's one big abstraction |
| 21:21.59 | [g2] | it's just a IXP425 part |
| 21:22.09 | [g2] | with a given rev level |
| 21:22.10 | beewoolie | And your only connection to it is JTAG? |
| 21:22.28 | [g2] | no JTAG is just one view |
| 21:22.42 | beewoolie | Then I don't follow. |
| 21:23.01 | beewoolie | From what I've seen so far, I don't see why you couldn't use a spreadsheet. |
| 21:23.56 | [g2] | you can cut and paste spread sheet in scripts |
| 21:24.09 | [g2] | beewoolie are you familiar with WinView at all ? |
| 21:24.19 | beewoolie | Nope. |
| 21:24.37 | [g2] | They've got a scripting capability |
| 21:24.54 | [g2] | so basically you can attach the JTAG connector and run the script |
| 21:25.03 | [g2] | the script does stuff like |
| 21:25.20 | [g2] | proc.cp15.foo=0x.... |
| 21:25.34 | [g2] | proc.memctrl.cs0=0x.... |
| 21:25.48 | [g2] | and all the names and values can be aliased |
| 21:26.03 | [g2] | it can also dump the values out |
| 21:26.34 | [g2] | so you can set them up and play with them in the GUI then save the snapshot of the proper values |
| 21:26.53 | [g2] | does that make any sense ? |
| 21:27.40 | [g2] | maybe the tools named WindICE |
| 21:28.45 | beewoolie | So, the idea would be to allow any serial text interface to control a remote JTAG box...or whatever. |
| 21:28.45 | beewoolie | It makes a little sense, though I don't see where you are going. |
| 21:28.45 | beewoolie | You could write this particular tool to use a BDI2000 for the time being. It's just a front-end on Expect. |
| 21:28.45 | beewoolie | perhaps with a little regex parsing built in. |
| 21:29.07 | [g2] | well the backend could be JTAG for board bring up |
| 21:29.13 | [g2] | or live SW on a running board |
| 21:29.28 | [g2] | just mmap the structures and pull / push values |
| 21:29.33 | beewoolie | These sorts of solutions are prone to failure given that the UI designers don't care to keep their output consistent. |
| 21:29.50 | beewoolie | Sure. That is a reasonable way to use the JTAG hardwar. |
| 21:30.08 | [g2] | well the UI should be driven by the meta language and the understanding of the part |
| 21:30.09 | beewoolie | It would make more sense to regularize the access methods so that there is on-going backward compatibility. |
| 21:30.20 | beewoolie | Easier said than done. |
| 21:31.01 | beewoolie | Breakage at this level is very frustrating to users who live in the M$ dialog boxes and scroll-bar universe. |
| 21:31.12 | beewoolie | All they see is that their wookey is broken. |
| 21:31.28 | [g2] | this is all part specific |
| 21:31.36 | beewoolie | They don't understand regex's and they eschew knowledge of command lines. |
| 21:31.55 | beewoolie | Actually, it's specific to the way that the information is gathered. |
| 21:32.15 | beewoolie | If you are engineering both the UI and the JTAG port, then there is no problem. In fact, a text parser isn't really the right solution. |
| 21:32.37 | beewoolie | If you control the backend, it is more efficient and reliable to compose a simple protocol. |
| 21:32.47 | beewoolie | SNMP is one such protocol, though I am not recommending it. |
| 21:32.53 | beewoolie | key=value. |
| 21:33.04 | beewoolie | key? returns the value. |
| 21:33.20 | beewoolie | Then it's a matter of defining a key space as well as value encodings. |
| 21:33.40 | [g2] | I'm quite familair with SNMP and that's totally the wrong place to go |
| 21:34.04 | [g2] | key=value is the same this as proc.mm.cs0=foo |
| 21:35.23 | beewoolie | Right. But you were talking about JTAG being only one backend. |
| 21:35.34 | ka6sox-office | BDM |
| 21:35.39 | ka6sox-office | used on PPC |
| 21:35.49 | vmaster | okay, then name the backend "debugger" |
| 21:35.55 | beewoolie | I guess I don't understand what the question is. |
| 21:35.59 | ka6sox-office | ya |
| 21:36.01 | vmaster | whatever the low-end connection is |
| 21:36.23 | ka6sox-office | JTAG/BDM/BDI whatever they want to call it. |
| 21:37.12 | beewoolie | Once you start with attempting to support these other devices, you run afoul with compatibility. |
| 21:37.13 | vmaster | is there another use of this, aside from debugging? |
| 21:38.31 | ka6sox-office | for BDM/BDI not really |
| 21:38.37 | ka6sox-office | JTAG of course |
| 21:38.52 | vmaster | mhh, no, "this" was refering to g2's idea |
| 21:39.00 | ka6sox-office | oh. |
| 21:39.15 | vmaster | an abstract notation of target "registers" |
| 21:39.20 | ka6sox-office | verification of a manufactured board. |
| 21:40.17 | vmaster | okay, but that again would be in the context of debugging - you somehow halt the target, put it into a debug mode, and ask it about the content of these registers, or set some of them |
| 21:40.50 | *** part/#openjtag NAiL (n=repvik@nslu2-linux/pdpc.active.repvik) |
| 21:41.03 | [g2] | vmaster you could do some of this live |
| 21:41.20 | ka6sox-office | with BDM/BDI the target is running |
| 21:41.22 | vmaster | heh, yeah, you could, but is it also something you would do |
| 21:41.30 | vmaster | okay, leave it running |
| 21:41.38 | vmaster | but it's outside of the normal use of the target |
| 21:41.39 | [g2] | also for automated vector generation intimate knowledge of the parts/layout is required |
| 21:42.33 | [g2] | vmaster in a PPC environment it's quite common to have the BDM/JTAG cable _always_ connected to the target |
| 21:43.28 | [g2] | at one company we put OS abstractions of the C++ kernel into the debugger |
| 21:43.46 | [g2] | allowing the debugger to understand the OS processes and runtime |
| 21:44.09 | vmaster | ah, lauterbach offers something like that, right? |
| 21:44.23 | [g2] | dunno, this was in-house |
| 21:45.02 | [g2] | s/scratch/from scratch/ |
| 21:45.33 | [g2] | hmm I guess purl couldn't handle that one |
| 21:45.46 | ka6sox-office | doesnt' seem like it |
| 21:47.34 | *** join/#openjtag [g2-lap] (n=g2@cpe-066-057-008-035.nc.res.rr.com) |
| 21:47.55 | [g2-lap] | he's a little peek at what I was talking about http://www.windriver.com/products/development_tools/debuggers_emulators/visionclick/visionCLICK_DS.pdf |
| 21:48.13 | [g2] | note to comment about the "record and playback" |
| 21:48.27 | [g2] | and different pictures of the internal views |
| 23:53.18 | [g2] | beewoolie ping |
| 23:53.27 | beewoolie | [g2]: yo |
| 23:53.43 | [g2] | you were playing with IDE drivers a little while ago right ? |
| 23:54.01 | beewoolie | I've written a few |
| 23:54.04 | [g2] | s/playing/wrote/ |
| 23:54.34 | [g2] | I've got an interesting situation with the CF/IDE driver for the Loft |
| 23:55.11 | beewoolie | In the Linux kernel? |
| 23:55.11 | [g2] | yeah |
| 23:55.11 | [g2] | there's a little patch |
| 23:55.17 | [g2] | but my problem now is in the kernel |
| 23:55.36 | beewoolie | How do you know you have a problem? |
| 23:55.45 | [g2] | here's now |
| 23:55.49 | [g2] | here's how |
| 23:56.21 | [g2] | For BE it works fine with the IDE probe but the data was byte swapped |
| 23:56.46 | [g2] | so a simple byte swapping of the 16Byte data fixes the problem for BE and all is well |
| 23:56.58 | [g2] | I can boot to the device, mount it all is happy |
| 23:57.21 | [g2] | However, in LE the device gets an error on the IDE probe |
| 23:57.31 | [g2] | same hardware, same kernel, just LE |
| 23:57.38 | [g2] | same sw |
| 23:57.40 | beewoolie | Sounds like the 'patch' is broken. |
| 23:57.59 | [g2] | this is in the generic part of the kernel |
| 23:58.13 | [g2] | there's no patch there |
| 23:58.17 | beewoolie | All IDE drivers go through a low-level portion that is target specific. |
| 23:58.28 | [g2] | right |
| 23:58.40 | [g2] | the functions are setup by the init |
| 23:58.55 | beewoolie | Is this all on 2.6.15? |
| 23:58.59 | [g2] | yep |
| 23:59.31 | beewoolie | What is the defconfig? |
| 23:59.51 | [g2] | I can e-mail or pastebin |
| 23:59.59 | beewoolie | Pastebin is OK. |