**** BEGIN LOGGING AT Sat Feb 11 10:59:56 2006
Feb 11 14:59:28 <ka6sox-web> morning
Feb 11 14:59:41 <AchiestDragon> hi
Feb 11 15:00:14 <ka6sox-web> howdy
Feb 11 15:00:24 <ka6sox-web> here at SCALE now
Feb 11 15:00:40 <AchiestDragon> kool
Feb 11 15:01:16 <ka6sox-web> gonna be fun but very busy.
Feb 11 15:01:32 <AchiestDragon> k
Feb 11 15:01:55 <ka6sox-web> I'll be online today from my lappy and will check in ;)
Feb 11 15:02:05 <AchiestDragon> :)
Feb 11 15:03:42 <ka6sox-web> cya after a bit!
Feb 11 15:04:01 <AchiestDragon> k
Feb 11 17:32:28 <lennert> i'm here
Feb 11 18:11:21 <AchiestDragon> question , tms,tck, tdi ,tdo and if used rtck need to run at full speed , but do   jtag1,2 and 3 ? and nTRST  need clocking at the same rate ??
Feb 11 18:15:00 <AchiestDragon> and second question are they required to be in sync with the jtag timing , from what i can see from the data these are indipendent lines that are ether in a stable state during jtag opperations or are used to initiate a feature
Feb 11 18:15:36 <vmaster> hmm?
Feb 11 18:15:48 <AchiestDragon> so there state is not likely to be changed half way though a jtag burst read 
Feb 11 18:15:56 <vmaster> jtag1, 2, 3?
Feb 11 18:16:23 <AchiestDragon> on the ep1200 board 
Feb 11 18:16:36 <AchiestDragon> 1 min
Feb 11 18:17:38 <vmaster> jtag1 is rtck
Feb 11 18:18:03 <AchiestDragon> yes
Feb 11 18:18:15 <vmaster> not sure what 2 and 3 are used for
Feb 11 18:18:34 <vmaster> they're neither specified by jtag nor by arm micros for example
Feb 11 18:18:55 <vmaster> ah, could be that these are dbgrq and dbgack
Feb 11 18:19:01 <AchiestDragon> on the 20 pin arm   jtag 2 would be  bgreq , and 
Feb 11 18:19:03 <vmaster> but this isn't used anywhere
Feb 11 18:19:04 <AchiestDragon> yes
Feb 11 18:19:23 <AchiestDragon> k , but there fitted here 
Feb 11 18:19:24 <vmaster> if they're used they're asynchronous
Feb 11 18:20:03 <AchiestDragon> but there use dose not relate to the normal jtag signal streem 
Feb 11 18:20:09 <vmaster> no
Feb 11 18:20:19 <AchiestDragon> like  nRESET  
Feb 11 18:20:22 <vmaster> yep
Feb 11 18:20:36 <AchiestDragon> that is indipendent 
Feb 11 18:24:26 <AchiestDragon> if we need to clock thease at the jtag clock rate then s be it , but if there seperate from it its easyer to sort out a jtag speed dubbler in the head  
Feb 11 18:32:52 <vmaster> they're independent, and don't have to be clocked at the jtag clock rate
Feb 11 18:34:27 <AchiestDragon> kool,, thats for the arm pinout ,, what about  emu0 and emu1 on the ti/tms jtag pinout  ?
Feb 11 19:18:47 <beewoolie-afk> ka6sox-office: ping?
Feb 11 19:29:30 <AchiestDragon> ka6sox-office:  is ... <ka6sox-web> here at SCALE now
Feb 11 19:30:01 <AchiestDragon> may be on for a bit later
Feb 11 20:23:17 <beewoolie-afk> Right.  I remember that he was going to that.
Feb 11 20:23:24 <beewoolie-afk> thx.
Feb 11 20:23:41 <beewoolie-afk> AchiestDragon: Are you at SCALE, too?
Feb 11 20:24:03 <AchiestDragon> no
Feb 11 20:24:10 <AchiestDragon> bit far for me 
Feb 11 20:24:15 <beewoolie-afk> That's what I thought.
Feb 11 20:24:32 <beewoolie-afk> I was wondering about the way we plan to set the JTAG TCK clock speed.
Feb 11 20:24:51 <beewoolie-afk> In the ATP protocol, I left this somewhat vague since I don't know how best to encode it.
Feb 11 20:24:54 <AchiestDragon> a dds chip 
Feb 11 20:25:05 <beewoolie-afk> That's what ka6sox wrote earlier.
Feb 11 20:25:17 <beewoolie-afk> What does that mean for configuration?  Do we set it with an integer frequency?
Feb 11 20:25:42 <AchiestDragon> 1 min
Feb 11 20:26:31 <AchiestDragon> the dds provides a frquancy from 0 to 50 mhz 
Feb 11 20:26:42 <beewoolie-afk> what is the smallest step?
Feb 11 20:26:46 <AchiestDragon> where multiplying this in the fpga
Feb 11 20:27:06 <AchiestDragon> but its about 1hz or 10hz 
Feb 11 20:27:33 <beewoolie-afk> so we could let the config set the frequency in Hz and round down to the closest acceptable value.
Feb 11 20:27:35 <AchiestDragon> the frequancy is set by sending it setup inof on a spi
Feb 11 20:27:51 <beewoolie-afk> OK.  I'm looking to solidify the protocol.
Feb 11 20:28:18 <beewoolie-afk> It looks like a 32 bit number carrying the frequency will be the most appropriate.
Feb 11 20:28:18 <AchiestDragon> AD9834  chip from analog devices 
Feb 11 20:30:13 <AchiestDragon> the chip needs 28 bits 
Feb 11 20:30:57 <AchiestDragon> and its 0 to 25mhz output  with a .2hz step 
Feb 11 20:31:20 <AchiestDragon> more than adiquate
Feb 11 20:33:07 <beewoolie-afk> That isn't the question.
Feb 11 20:33:23 <beewoolie-afk> If we are going to set a protocol element for this, we need to make sure that it makes sense.
Feb 11 20:34:48 <beewoolie-afk> It looks like we can let it be a frequency.  A 32 bit number is sufficient although we won't be able to reach the slowest frequency.
Feb 11 20:35:29 <beewoolie-afk> I think we can be OK with 1Hz as the slowest, but we could allow for four bits of a fraction.
Feb 11 20:35:42 <beewoolie-afk> That would give us a 268MHz peak frequency.
Feb 11 20:36:29 <AchiestDragon> k ,, the freq that isset by the dds is fraction of the actual feq anyway 
Feb 11 20:36:50 <beewoolie-afk> Right.  But again, that isn't the issue.
Feb 11 20:37:01 <beewoolie-afk> It doesn't matter how it's implemented.
Feb 11 20:37:06 <AchiestDragon> as we have programable multiplication of the frequancy   *2  , *4 or *8
Feb 11 20:37:13 <beewoolie-afk> what I'm talking about is what the user can set as the frequency.
Feb 11 20:37:31 <beewoolie-afk> It is somewhat arbitrary, but we do want to make a reasonablechoice.
Feb 11 20:38:00 <beewoolie-afk> There are two constraints.  It is not good if our encoding prevents users from setting useful frequencies.
Feb 11 20:38:05 <AchiestDragon> then maybe better to have it store the frequancy in  xxx.xxx Mhz 
Feb 11 20:38:08 <beewoolie-afk> We're unlikely to make this choice.
Feb 11 20:38:31 <beewoolie-afk> What if the user wants to set a 10Hz JTAG clock?
Feb 11 20:38:47 <AchiestDragon> there is no need for a fine scale than that
Feb 11 20:38:51 <AchiestDragon> that slow /
Feb 11 20:38:53 <beewoolie-afk> The question is not so much how it is implemented but what range of frequencies do we want to offer.
Feb 11 20:38:53 <AchiestDragon> ?
Feb 11 20:39:07 <beewoolie-afk> So, the slowest is 1KHz.
Feb 11 20:39:22 <AchiestDragon> that should be fine 
Feb 11 20:39:57 <beewoolie-afk> The we can get away with a 16 bit count of KHz for the frequency.
Feb 11 20:40:00 <AchiestDragon> the top beeing 999Mhz is over what we can do atm , but in 5 years  who knows  
Feb 11 20:40:15 <beewoolie-afk> It isn't BCD.  It's binary.
Feb 11 20:40:23 <AchiestDragon> yes 
Feb 11 20:40:24 <beewoolie-afk> So, the limit would be 65MHz.
Feb 11 20:40:37 <beewoolie-afk> Which matches the design.
Feb 11 20:40:45 <AchiestDragon> were looking at 300mhz
Feb 11 20:40:59 <AchiestDragon> as the system speed not the osc speed 
Feb 11 20:41:11 <beewoolie-afk> the user only cares about the TCK speed.
Feb 11 20:41:30 <AchiestDragon> yes a 300mhz tck 
Feb 11 20:42:08 <beewoolie-afk> OK.  Then maybe it should be a 16 bit count of 10KHz.  That's 655 MHz
Feb 11 20:42:48 <AchiestDragon> k kool ,
Feb 11 20:42:52 <beewoolie-afk> THX
Feb 11 20:43:10 <beewoolie-afk> I may bring it down to a count of 5KHz so that we can reach a little lower.
Feb 11 20:43:59 <AchiestDragon> think in the future higher may be better ,,,  
Feb 11 20:44:13 <beewoolie-afk> But this design isn't going to go faster than that.
Feb 11 20:44:33 <beewoolie-afk> The idea is to make the protocol stable for this design.
Feb 11 20:44:47 <beewoolie-afk> Modifying it for a different encoding is reasonable in a different design.
Feb 11 20:45:14 <beewoolie-afk> Heck, I suppose we could just go for a 32 bit number and forget about it.
Feb 11 20:45:19 <AchiestDragon> the head is the limiting factor atm because the chips we can get limit that to 375mhz top , but the rest could drive the head to 600 if the head couldrun at that 
Feb 11 20:45:20 <beewoolie-afk> It's not like we're saving anything.
Feb 11 20:45:57 <AchiestDragon> so in 1 or 2 years when there are faster chips it could do that 
Feb 11 20:46:23 <beewoolie-afk> OK.  I'll just leave it at a simple 32 bit frequency.  That gives us plenty of room for faster designs.  4.2 GHz.
Feb 11 20:48:07 <AchiestDragon> looking at the specs of some of the cpu's there jtag is 1/16th of the cpu clock so a 4.7Ghz cpu can have a 293.75Mhz tclk
Feb 11 20:49:09 <AchiestDragon> when they get a 9.4hz cpu thats 587.5Mhz ,, possabaly 2 to 4 years away 
Feb 11 20:49:26 <beewoolie-afk> This can be a tough call.  using a 32 bit number is probably over the top, but it does make it clear.  Alternatively, we could encode the frequency in a field saving room for expansion without requring a change to the protocol.
Feb 11 20:49:42 <beewoolie-afk> I think there is a good reason we haven't gotten CPUs that fast.
Feb 11 20:49:57 <beewoolie-afk> I've read that they aren't planning to go that fast.  It's cheaper to put in more cores.
Feb 11 20:50:24 <beewoolie-afk> CPU clock frequencies have been kinda static for some years.
Feb 11 20:50:35 <AchiestDragon> maybe  ,but  4.7Ghz is already avalable on some , its the intel chipset that has maxed out 
Feb 11 20:50:56 <AchiestDragon> the die size for it is its limit 
Feb 11 20:50:57 <beewoolie-afk> I'm OK using a 32 bit frequency value.
Feb 11 20:51:18 <beewoolie-afk> We aren't likely to encounter problems and we don't need to conserve two bytes.
Feb 11 20:52:14 <AchiestDragon> that should allow for a 0 to 4.2G range at 1hz steps    yes 
Feb 11 20:52:32 <beewoolie-afk> Zero is reserved for adaptive clocking.
Feb 11 20:53:16 <AchiestDragon> yes , and there nees to be a upper an lower limit 
Feb 11 20:54:01 <AchiestDragon> defineable somewhare as a constant or varible , dependent on the hardware 
Feb 11 20:54:25 <beewoolie-afk> That is another issue.
Feb 11 20:54:39 <beewoolie-afk> There are a couple of ways of handling this.
Feb 11 20:54:54 <beewoolie-afk> We can implement a query function so that the host can find out what frequencies are available.
Feb 11 20:55:08 <AchiestDragon> im looking at trying to auto detect the head atached also somehow 
Feb 11 20:55:10 <beewoolie-afk> Or, we can simple round down when we get a very high frequency.
Feb 11 20:55:35 <beewoolie-afk> I'm not worried about this ATM.  It doesn't have to affect the protocol.
Feb 11 20:56:03 <AchiestDragon> true , that can be sorted in the aplication 
Feb 11 20:56:33 <beewoolie-afk> The BSDL files define limits on TCK.
Feb 11 20:57:16 <AchiestDragon> and if over the system  speed then the system just runs as fast as it can 
Feb 11 20:57:39 <beewoolie-afk> Indeed.
Feb 11 22:42:09 <[g2]> hey beewoolie-afk 
Feb 12 00:47:38 <beewoolie-afk> [g2]: Hey.  Been busy making desert.
Feb 12 00:52:14 <[g2]> hope it's a yummy one -- JTAG pie!
Feb 12 00:54:20 <beewoolie-afk> Creme Brulee again.
Feb 12 01:10:51 <[g2]> got a good recipe ?
Feb 12 01:11:07 <[g2]> if so you'll have to e-mail me :)
Feb 12 06:19:55 <ka6sox> its waay too early
Feb 12 07:41:33 <velinp> ~seen lennert
Feb 12 07:41:36 <purl> lennert is currently on #openjtag (9d 8m 39s). Has said a total of 98 messages. Is idling for 14h 9m 12s, last said: 'i'm here'.
Feb 12 07:43:33 <ka6sox> ~seen velinp
Feb 12 07:43:41 <purl> velinp is currently on #openjtag (58m 47s). Has said a total of 1 messages. Is idling for 2m 7s, last said: '~seen lennert'.
Feb 12 07:43:45 <velinp> ka6sox: hi
Feb 12 07:43:49 <ka6sox> howdy
Feb 12 07:44:17 <ka6sox> still here in Lost Angles.
Feb 12 07:44:53 <velinp> ka6sox: not happy in LA? 
Feb 12 07:45:06 <ka6sox> I moved out of here in 1989
Feb 12 07:45:19 <ka6sox> its not my favourite place.
Feb 12 07:46:03 <ka6sox> I haven't seen Lennert in a while.
Feb 12 07:46:18 <velinp> purl says 14 hours
Feb 12 07:46:37 <velinp> about Lennert; 
Feb 12 07:48:06 <velinp> I am looking for him, too. I had a buildd problem with mingw32 (-binutils); ld caused _many_ alignment errors; fixed by 3 > /proc/cpu/alignment; wanted to ask if it's a right solution 
Feb 12 07:49:39 <ka6sox> k
Feb 12 07:52:59 <velinp> mingw32 build finished; wanted to go back 'echo 1>/proc/cpu/alignment', but it stays at 3 (warn + fix), doesn't warn either; funny ...
Feb 12 08:01:57 <velinp> sorry; maybe i misspelled something; did set alignment back to 1 (default)
Feb 12 08:03:14 <ka6sox> ah
Feb 12 08:27:45 <lennert> i'm here
Feb 12 08:27:56 <lennert> i was intoxicated yesterday so wasn't online much
Feb 12 08:29:18 <velinp> lennert: hi; you able to answer questions now :) ?
Feb 12 08:29:44 <lennert> yeah :)
Feb 12 08:31:05 <velinp> i had a buildd problem with mingw32 (-binutils); ...-ld was generating _lots_ of alignment errors; fixed by 3 -> /proc/cpu/alignment; q: what is the right thing for arm 1 or 3 ?
Feb 12 08:31:36 <lennert> i have '0' in there
Feb 12 08:31:48 <lennert> # grep faults /proc/cpu/alignment
Feb 12 08:31:50 <lennert> User faults:    0 (ignored)
Feb 12 08:32:14 <lennert> if there's userland code that causes alignment faults, that would seem like a bug to me
Feb 12 08:32:29 <velinp> i would prefer warn + fix on a buildd; what do you say ?
Feb 12 08:33:25 <lennert> no
Feb 12 08:33:38 <lennert> there might be applications that test in their configure scripts whether unaligned accesses are OK
Feb 12 08:33:52 <lennert> if you have 'fix' on, you might end up with a binary that doesn't work if you have fix off
Feb 12 08:34:00 <lennert> 'warn' is okay to turn on, but i wouldn't turn 'fix' on
Feb 12 08:34:28 <lennert> does that make sense?
Feb 12 08:34:50 <velinp> so, you reccomend 1 == warn (+ regular monitoring :) )
Feb 12 08:34:57 <lennert> yeah :)
Feb 12 08:35:14 <velinp> ok; thanks; how is the 2-cpu beast doing ?
Feb 12 08:53:46 <lennert> so, 2.6 boots on it, and i've been trying to get the ethernet going
Feb 12 08:56:12 <velinp> good luck with it :) ka6sox was asking about you ~1 hour ago from 'Lost Angles'; bye from me :)
Feb 12 08:56:20 <lennert> no ETA on that (the 'how to initialise the networking' document is 50 pages alone)
Feb 12 08:56:23 <lennert> thanks :)
Feb 12 08:56:55 <lennert> i read the backlog
Feb 12 08:56:55 <ka6sox> lennert, you are famous...I just say lennert and everyone nods.
Feb 12 08:56:58 <lennert> i hope he's around
Feb 12 08:57:03 <lennert> ka6sox: haha
Feb 12 08:58:33 <lennert> lots of big names on the speaker list
Feb 12 08:58:44 <ka6sox> ya
Feb 12 08:58:59 <ka6sox> john terpstra came by today...had a good laugh with him.
Feb 12 08:59:55 <lennert> ah, samba folk
Feb 12 09:00:25 <ka6sox> ya...he came by and told me that "Samba Sucks!"
Feb 12 09:00:38 <lennert> samba4 a bit less than the previous versions, though
Feb 12 09:00:51 <lennert> i do hope they get a release out soon
Feb 12 09:01:32 <lennert> http://www.desktoplinux.com/articles/AT6464009101.html
Feb 12 09:01:38 <lennert> "SCO Evangelist John Terpstra on Windows Pains"
Feb 12 09:01:40 <ka6sox> I think that they will soon.
Feb 12 09:01:45 <lennert> the beast!
Feb 12 09:14:44 <AchiestDragon> bbl , out for a bit 
Feb 12 09:18:29 <ka6sox> bya..I'm going to bed.
Feb 12 09:18:43 <lennert> ka6sox: g'nite
**** ENDING LOGGING AT Sun Feb 12 10:59:58 2006