**** BEGIN LOGGING AT Thu Sep 22 02:59:56 2005
Sep 22 06:33:11 <prpplague> anyone here have experience loading apps into ram via jtag to an arm device?
Sep 22 06:48:54 <vmaster_> yes
Sep 22 06:49:38 <ka6sox> beewoolie does
Sep 22 06:50:15 <[g2]> ep1220 has into the mini I cache
Sep 22 06:50:56 <prpplague> any tips/tricks i need to know about?
Sep 22 06:50:57 <[g2]> we were talking about loading a small program and setting up the memory controller and serial as bootstrap
Sep 22 06:51:14 <vmaster> prpplague: what arm?
Sep 22 06:51:15 <ep1220> mini cache is XScale specific
Sep 22 06:51:26 <prpplague> vmaster: arm7tdmi core
Sep 22 06:51:40 <prpplague> vmaster: s3c44b0x to be exact
Sep 22 06:51:42 <vmaster> openocd.berlios.de - read my thesis, i hope it explains everything necessary
Sep 22 06:51:54 <prpplague> vmaster: oh lovely, thanks
Sep 22 06:51:58 <prpplague> vmaster: 
Sep 22 06:52:10 <prpplague> vmaster: docs are always helpful :)
Sep 22 06:52:21 <vmaster> the code from july works for arm7 and arm9, but i only tested it on at91rm9200 and hynix hms307202
Sep 22 06:52:58 <vmaster> i gotta go - cya
Sep 22 06:53:03 <prpplague> vmaster: ok thanks
Sep 22 06:54:09 <prpplague> [g2]: yea thats what we are talking about
Sep 22 06:55:01 <prpplague> [g2]: i want to setup something similiar Haret
Sep 22 06:57:55 <prpplague> [g2]: just not wince specific
Sep 22 07:01:33 <[g2]> it's a common thing to want to do
Sep 22 07:01:44 <[g2]> manufacturing turn up
Sep 22 07:06:27 <prpplague> [g2]: well this is a little different from what i would normally do
Sep 22 07:06:37 <prpplague> [g2]: normally i would just load the flash with a bootloader
Sep 22 07:06:57 <prpplague> [g2]: in this case i want to load directly to ram and execute from there
Sep 22 07:06:59 <[g2]> bootloader or test sw, same deal really
Sep 22 07:07:17 <prpplague> [g2]: yea, without the interupt vectors though 
Sep 22 07:07:50 <prpplague> [g2]: yea i figured it was pretty common deal, just hadn't done it before
Sep 22 07:07:51 <[g2]> there's nothing wrong with polled sw, it's been run for years
Sep 22 07:08:09 <[g2]> :)
Sep 22 07:09:10 <prpplague> [g2]: i'm looking ot write something with small serial cli that i can peek/poke adresses and do some mem location dumps
Sep 22 07:09:27 <prpplague> [g2]: i probably could use most of the code from some small bootloader
Sep 22 07:11:11 <[g2]> prpplague, we should consider a protocol (serial first) that dumps information out
Sep 22 07:11:34 <[g2]> VxWorks VisionICE iirc had a hw descriptor file thingy
Sep 22 07:11:53 <[g2]> I've been wanting to setup something similar for IXP4XX
Sep 22 07:12:22 <[g2]> connect, get state, have a GUI display _everything_ 
Sep 22 07:12:37 <[g2]> and then have it update able
Sep 22 07:18:15 <prpplague> [g2]: hmm, i was just thinking more like small cli terminal that could be accessed via minicom or hyperterminal
Sep 22 07:18:35 <prpplague> [g2]: but a good serial protocol would work as well, just would need some sort of client to handle the data
Sep 22 07:19:17 <[g2]> prpplague exactly
Sep 22 07:19:27 <[g2]> and some tools to generate it
Sep 22 07:19:47 <prpplague> hmm
Sep 22 07:19:56 <[g2]> possibly a DB where ppl could copy part info
Sep 22 07:23:06 <[g2]> bbl
Sep 22 07:24:24 <prpplague> [g2]: yea, later
Sep 22 09:11:01 <[g2]> ep1220 ping 
Sep 22 09:11:08 <[g2]> bullet welcome
Sep 22 09:11:14 <bullet> hi :)
Sep 22 09:11:33 <[g2]> ep1220 is the guy that did the JTAG to the mini I cache and is building the hw
Sep 22 09:11:43 <bullet> k
Sep 22 09:12:17 <[g2]> ep1220 How long do you think it'd take to JTAG the entire I cache ?
Sep 22 09:12:49 <[g2]> bullet I think a small version of APEX will fit into the I cache, we 
Sep 22 09:13:02 <[g2]> would just need to init memory and the serial
Sep 22 09:13:08 <[g2]> and could run from RAM
Sep 22 09:15:35 <bullet> ui, I cache is only 32kb. nice
Sep 22 09:37:02 <vmaster> uhm, you can't use the icache like normal ram, afaik
Sep 22 09:37:25 <vmaster> it's used only for instruction fetches
Sep 22 09:52:59 <ep1220> [g2]: I think You can get 2 or 3MByte/sec to the i-chache
Sep 22 09:53:25 <ep1220> but as it is only 32K even 32KB/sec woudl fill it within a second :-)
Sep 22 09:54:16 <ep1220> sorry meant 2 or 3MBit/sec
Sep 22 09:58:23 <bullet> ep1220: your ftdi-jtag project looks very nice
Sep 22 09:59:27 <bullet> what about the software for using it?
Sep 22 10:13:00 <ep1220> bullet: ibb  in 2 hours
Sep 22 10:32:22 <[g2]> ep1220, thx.
Sep 22 10:33:19 <[g2]> Right so downloading a mini bootloader and then possiblly dl'ing an image could be pretty quick
Sep 22 12:10:04 <ep1220> back now
Sep 22 12:13:43 <bullet> so, how's the status on your project?
Sep 22 12:14:48 <ep1220> I have a prototype and core-SW to talk to the FTDI on Win and Linux
Sep 22 12:15:22 <ep1220> plus Sw to use this core to acess the XScale JTAG debug registers
Sep 22 12:15:24 <bullet> and the software is quite hardware specific, right?
Sep 22 12:16:00 <ep1220> it is for the FTDI2232C
Sep 22 12:16:14 <ep1220> or you mean target hardware ?
Sep 22 12:16:18 <bullet> i mean, target hardware, yes
Sep 22 12:16:47 <ep1220> the core is entirely target independent
Sep 22 12:17:10 <ep1220> the XSCALE part only works on a XSCALE CPU of course.
Sep 22 12:17:36 <ep1220> So far I did not write bootloaders,FLASH programmers, ..
Sep 22 12:19:31 <bullet> i'm actually new to this stuff, so i don't really know, what jtag can do. for me at the moment, writing the flash of nslu2 would be the primary target, so...
Sep 22 12:20:52 <ep1220> JTAG can do this.
Sep 22 12:21:05 <bullet> that at least i knew already :)
Sep 22 12:21:07 <ep1220> there are 2 ways:
Sep 22 12:21:37 <ep1220> 1) You use the scan-chain. This is the way jtagtools work.
Sep 22 12:21:53 <ep1220> Need explanation of "scan-chain" ?
Sep 22 12:22:02 <bullet> i'll look it up
Sep 22 12:22:27 <ep1220> good. Disadvantage: you have to move a lot of bits, so it is slow
Sep 22 12:22:45 <ep1220> Advantage: it is "realtive" independent of the used CPU.
Sep 22 12:23:08 <ep1220> You just need to tell the soft the position of the address,data and some control lines.
Sep 22 12:23:25 <ep1220> 2) You use the CPU to prog the Flash
Sep 22 12:23:52 <ep1220> here you have to write Flash programming software for (in your case) the Xscale.
Sep 22 12:24:08 <bullet> and this software would run "on" the xscale itself?
Sep 22 12:24:11 <ep1220> You have no bootloader or operating system at this point.
Sep 22 12:24:28 <ep1220> Yes, it would be copied into the Xscale I-cache.
Sep 22 12:25:01 <ep1220> it then can read data from JTAG or any other interface for which you include a driver
Sep 22 12:25:26 <ep1220> JTAG is the least work (just poll a ready bit and do a read)
Sep 22 12:26:36 <bullet> hmm, and the scan-chain method, does it work with your software already?
Sep 22 12:26:44 <ep1220> no
Sep 22 12:27:11 <bullet> so, you actually don't use it at all for flash-programming yet?
Sep 22 12:27:15 <ep1220> yes
Sep 22 12:28:04 <ep1220> my primary interest is debugging.
Sep 22 12:28:32 <ep1220> Do you want to flash just once ?
Sep 22 12:28:47 <vmaster> ep1220: your jtag core software - do you queue the scan commands? or do you send them whenever they're needed?
Sep 22 12:28:54 <bullet> no, well. it would be for development. so, i don't know yet :)
Sep 22 12:29:32 <bullet> i might assemble one of those primitive parallel port thingies, would suffice for me, i think.
Sep 22 12:29:53 <bullet> at least, for now
Sep 22 12:29:55 <ep1220> unless You plan to develop a bootloader, this is likely the case.
Sep 22 12:30:16 <[g2]> ep1220, boards ready yet ?
Sep 22 12:30:20 <ep1220> i heard it takes 40 minutes to flash redboot with such a device
Sep 22 12:30:35 <bullet> yeah, heard something like that too :(
Sep 22 12:30:48 <ep1220> [g2]: no, still discussing features with ka6sox 
Sep 22 12:30:59 <ep1220> vmaster: the commands are queued
Sep 22 12:31:16 <vmaster> is that software available somewhere?
Sep 22 12:31:24 <[g2]> ep1220 how do I get a 14 and 20 pin proto for Xscale ?
Sep 22 12:31:25 <[g2]> :)
Sep 22 12:31:50 <ep1220> [g2]: we plan to make adapters for several JTAG IFs.
Sep 22 12:31:53 <bullet> ep1220: are you planing on selling assembled boards?
Sep 22 12:33:10 <ep1220> [g2]: one of these will be ARM 20pin to ARM14pin.
Sep 22 12:34:15 <ep1220> bullet: At first we plan to make a few boards, they use thru-hole technology except for 2 smd (SO-16)
Sep 22 12:35:07 <ep1220> bullet: i might solder a few for people planing to help write software ;-)
Sep 22 12:35:28 <bullet> hehe
Sep 22 12:35:47 <ep1220> only after there is usefull  software it makes sense to think of a production run
Sep 22 12:36:02 <bullet> yeah
Sep 22 12:37:49 <vmaster> if you're interested have a look at this: http://www.fh-augsburg.de/~hhoegl/proj/volksmikro/usb-jtag/050910/ - the new version of the board my professor designed - he's using galvanically isolated level shifters to fulfill some obscure requirement he has for another project
Sep 22 12:37:55 <bullet> at least you seem to be just a few kilometers away ;)
Sep 22 12:39:05 <ep1220> I see, You live in our western neighbour
Sep 22 12:39:33 <bullet> :)
Sep 22 12:40:58 <bullet> vmaster: but other than that, it's quite the same, eh?
Sep 22 12:41:00 <ep1220> vmaster: Thx. So far my SW is on no public server
Sep 22 12:43:48 <vmaster> it doesn't use the dlp2232m anymore, as that's quite expensive, at least from the distributors he queried
Sep 22 12:54:08 <ep1220> vmaster: Do You have SMD finepitch soldering equipment at the university ?
Sep 22 12:55:29 <vmaster> i think so - he's having the ee department solder the stuff for us
Sep 22 23:29:42 <ka6sox> morning
Sep 22 23:29:52 <ep1220> hello
Sep 22 23:30:25 <ka6sox> almost time for bed.
Sep 22 23:30:50 <ka6sox> but I was wondering if a switch or just jumpers would be better :)
Sep 22 23:31:30 <ep1220> i have no real preference.
Sep 22 23:31:50 <ep1220> jumper = cheaper, switch = less board space
Sep 22 23:32:33 <ka6sox> 6 of one, 1/2 dozen of the other.
Sep 22 23:33:03 <ep1220> i would try to use a rotary switch
Sep 22 23:33:04 <ka6sox> going to be easier to recreate (buy local parts) with just jumpers.
Sep 22 23:34:12 <ep1220>  You prefer jumper ?
Sep 22 23:34:38 <ka6sox> yes...because its changed infrequently
Sep 22 23:35:05 <ka6sox> and finding the "exact" switch locally might prove difficult.
Sep 22 23:35:27 <ep1220> ok
Sep 22 23:36:09 <ka6sox> I think that we are zeroing in on it!
Sep 22 23:36:30 <ka6sox> thats kind of the icing on the cake (being able to use all those voltages)
Sep 22 23:37:02 <ep1220> yeah, the low voltages are coming ...
Sep 22 23:37:40 <ka6sox> uh huh...I can't remember what the S3e is going to be...1.8v?
Sep 22 23:38:06 <ep1220> do not know. 
Sep 22 23:38:27 <ka6sox> okay...when you are ready I'll look again :)
Sep 22 23:38:32 <ep1220> one more thing: If we do away the serial we could use a DAC to set the voltage :-)
Sep 22 23:39:10 <ka6sox> the jumper is more reliable :)
Sep 22 23:39:21 <ka6sox> and cheaper.
Sep 22 23:39:38 <ka6sox> I suppose you could cause a problem.
Sep 22 23:39:56 <ka6sox> (either way by setting it wrong)
Sep 22 23:40:15 <ep1220>  also true for jumper.
Sep 22 23:40:22 <ka6sox> yes
Sep 22 23:40:32 <ka6sox> but you have a visual indication.
Sep 22 23:41:02 <ka6sox> how about using the parallel bits to set the voltage?
Sep 22 23:41:09 <ka6sox> thru a fet switch?
Sep 22 23:41:57 <ep1220> is also possible. 
Sep 22 23:42:16 <ka6sox> "there is more than one way to do it"
Sep 22 23:42:26 <ep1220> rrright
Sep 22 23:42:52 <ka6sox> :)
Sep 22 23:43:07 <ka6sox> okay I'm game for any of those ways..
Sep 22 23:43:17 <ka6sox> the dac is my least fave as we lose the serial.
Sep 22 23:45:08 <ep1220> I can ceck, maybe one can have the serial plus use the other b-ports bits independent of it.
Sep 22 23:45:25 <ka6sox> okay sounds good.
Sep 22 23:46:00 <ep1220> anyway, i plan to finalize the design by this evening
Sep 22 23:46:17 <ep1220> (else we might have more good ideas ;-)
**** ENDING LOGGING AT Fri Sep 23 02:59:56 2005