| 02:39.20 | *** join/#openjtag ByronT (~byron-por@ByronT.nslu2-linux) |
| 03:34.23 | *** join/#openjtag basstech (~basstech@ip68-111-62-90.br.no.cox.net) |
| 06:21.24 | jbowler-away | Am I missing something? That code seems to be writing to the physical address of the GPIO within the loop, but writing the GPIO is very slow - it won't show the core clock speed at all. |
| 06:23.49 | jbowler-away | mov r0, #0 |
| 06:23.58 | jbowler-away | 0: subs r0, r0, #1 |
| 06:24.01 | jbowler-away | bne 0b |
| 06:24.59 | jbowler-away | Should take about 32s with a 266MHz core clock (2^33 instructions) |
| 08:26.15 | *** join/#openjtag ep1220 (~NN@gate.epatec.at) |
| 08:26.35 | ep1220 | [g2]-away: ping |
| 08:27.34 | *** join/#openjtag VoodooZ_Log (~VoodooZ@CPE00c0f021712c-CM014110003785.cpe.net.cable.rogers.com) [NETSPLIT VICTIM] |
| 08:28.03 | *** join/#openjtag nslu2-log (dyoung@netblock-66-159-209-60.dslextreme.com) |
| 08:28.07 | *** join/#openjtag rwhitby-away (~rwhitby@CPE-60-226-247-226.sa.bigpond.net.au) |
| 08:29.19 | *** join/#openjtag ByronT-Away (~byron-por@adsl-212-196-167.bhm.bellsouth.net) [NETSPLIT VICTIM] |
| 08:29.19 | *** join/#openjtag Jacmet (~jacmet@50C59F0E.flatrate.dk) [NETSPLIT VICTIM] |
| 08:32.08 | *** join/#openjtag NAiL (~repvik@banan.dataparty-mn.biz) |
| 08:32.34 | *** join/#openjtag Jacmet (~jacmet@50C59F0E.flatrate.dk) |
| 08:32.48 | *** join/#openjtag VoodooZ_Log (~VoodooZ@CPE00c0f021712c-CM014110003785.cpe.net.cable.rogers.com) |
| 08:33.20 | *** join/#openjtag ByronT-Away (~byron-por@adsl-212-196-167.bhm.bellsouth.net) [NETSPLIT VICTIM] |
| 08:35.39 | *** join/#openjtag dyoung (~dyoung@cpe-24-94-88-62.hawaii.res.rr.com) [NETSPLIT VICTIM] |
| 08:43.33 | *** join/#openjtag ka6sox-office (~tking@netblock-66-159-209-52.dslextreme.com) |
| 09:18.56 | [g2]-away | ep1220, pong |
| 09:19.25 | ep1220 | hey; i ran the blink prog |
| 09:19.30 | [g2] | cool |
| 09:19.36 | [g2] | ep1220, IS THE MAN! |
| 09:19.38 | ep1220 | the frequency i see is 5MHz |
| 09:19.54 | ep1220 | so a lot of questions |
| 09:20.37 | ep1220 | e.g. are there waitStates on access to the GPIO area ? |
| 09:20.50 | [g2] | do you get a nice shape of the waiveform ? |
| 09:20.55 | [g2] | waveform |
| 09:20.57 | ep1220 | yes |
| 09:21.12 | [g2] | good |
| 09:21.34 | [g2] | my understanding was that the GPIO are off the north AMBA bus |
| 09:21.41 | [g2] | but I not certain |
| 09:22.15 | [g2] | hmm.. just waking up ... I am |
| 09:22.36 | ep1220 | :-) I am close to lunch |
| 09:22.48 | ep1220 | maybe we talk again in 1 hour ? |
| 09:22.49 | [g2] | 5:22 AM here :) |
| 09:23.18 | [g2] | If you are heading out to lunch that fine. |
| 09:23.30 | [g2] | I'm here now, I'll be around later |
| 09:23.40 | [g2] | THX for all the help |
| 09:24.00 | ep1220 | np; |
| 09:24.09 | [g2] | it's 2:23 for jacques and bewoolie :) |
| 09:24.22 | [g2] | I doubt they'll be around for a while |
| 09:24.52 | dyoung | [g2] is a robot. |
| 09:25.00 | [g2] | :) |
| 09:25.09 | [g2] | hey late sleeper :) |
| 09:25.13 | ep1220 | good, i'll be back in 1 hour |
| 09:25.24 | dyoung | I was looking at Lennerts slug. |
| 09:25.26 | [g2] | have a great lunch |
| 09:25.51 | [g2] | how's it look ? |
| 09:26.29 | dyoung | it looks dead. |
| 09:26.50 | dyoung | If I were him, and I had this slug, I would have thrown it out. |
| 09:32.14 | [g2] | dyoung, have you been following the slug "peformance" issue ? |
| 09:33.09 | dyoung | more or less. |
| 09:33.22 | dyoung | I was gonna go poking around for a clock halfer. |
| 09:40.05 | [g2] | the 5Mhz number is interesting |
| 09:43.17 | [g2] | dyoung, that's for looking at Lennert's flaky slug |
| 09:43.27 | [g2] | doh... THANKS for |
| 09:43.30 | dyoung | tnats? |
| 09:43.41 | dyoung | wow, I'm being punished with a 5Mhz slug? |
| 09:43.42 | dyoung | :-( |
| 09:44.15 | [g2] | well it's 15-20 Mhz |
| 09:44.30 | [g2] | at least 10 |
| 09:44.35 | dyoung | phew. |
| 09:44.35 | [g2] | xor and branch |
| 09:44.37 | dyoung | :-) |
| 09:44.48 | dyoung | I'm still digesting. |
| 09:45.04 | dyoung | I agree that these slugs are running well, sluggish. |
| 09:45.21 | dyoung | but I cant see how you make it run at half speed without screwing up the pci clock |
| 09:45.33 | [g2] | maybe after we started calling them "slug" the went on strike part time :) |
| 09:46.04 | [g2] | I'm with beewoolie.... Undocumented features |
| 09:46.20 | [g2] | Intel says .... "just do this...." |
| 09:47.10 | dyoung | hey, thats what I said. |
| 09:47.17 | dyoung | Undocumented fuse or soething |
| 09:47.32 | dyoung | bastage |
| 09:47.47 | dyoung | so linksys bought all the defective ixp420s |
| 09:48.00 | [g2] | I don't think so |
| 09:48.38 | dyoung | damn SX's |
| 09:49.02 | [g2] | maybe, but I'd guess they are clocked down for either heat (plastic case) or more likely product channel issues |
| 09:49.37 | [g2] | but I don't know |
| 09:49.52 | [g2] | I've been looking but haven't seen the RT-042 |
| 10:00.38 | NAiL | [g2]: 08:21 < jbowler-away> Am I missing something? That code seems to be writing to the physical address of the GPIO within the loop, but writing the GPIO is |
| 10:00.41 | NAiL | <PROTECTED> |
| 10:02.06 | [g2] | heh.... that was 2:21 here :) |
| 10:06.04 | NAiL | slugtime never sleeps ;) |
| 10:06.23 | dyoung | I have a serious headache. |
| 10:06.36 | NAiL | I've got painkillers? |
| 10:06.58 | NAiL | (and I've got bubblewrap) |
| 10:07.02 | dyoung | I try to avoid drugs. |
| 10:07.25 | NAiL | That is generally a good thing. |
| 10:07.25 | dyoung | NSAID dependancy sucks. |
| 10:07.32 | dyoung | (ibuprofen) |
| 10:07.37 | dyoung | Aka Vitamin I |
| 10:07.55 | dyoung | so does your bubblewrap have a PL2303 doodad in it? |
| 10:08.00 | NAiL | hmm |
| 10:08.15 | NAiL | I'll be damned |
| 10:08.21 | NAiL | There's something in here! |
| 10:08.30 | NAiL | WHO PUT STUFF IN MY BUBBLEWRAP? |
| 10:08.52 | dyoung | I think I sent Lennert my serial cable. |
| 10:08.56 | dyoung | USB/Serial/ |
| 10:09.02 | dyoung | thinking I was gonna make another one |
| 10:10.01 | NAiL | Sounds like you havent |
| 10:10.21 | NAiL | Ugh.. slow soldering iron |
| 10:25.14 | *** join/#openjtag cbpage (~cpage@ip68-6-39-177.sb.sd.cox.net) |
| 10:30.24 | *** join/#openjtag jacques (~jacques@c-67-171-198-193.hsd1.or.comcast.net) |
| 10:34.36 | *** join/#openjtag ChanServ (ChanServ@services.) |
| 10:34.36 | *** mode/#openjtag [+o ChanServ] by irc.freenode.net |
| 10:35.04 | NAiL | hmm... Why doesn't my slug boot when my serial cable is attached? |
| 10:35.24 | dyoung | what kind of serial cable? |
| 10:35.39 | NAiL | usb |
| 10:35.49 | dyoung | you might need the pullups |
| 10:36.56 | NAiL | It's the cable ka6sox sent |
| 10:37.02 | dyoung | do ou have some spare resistors lying around? |
| 10:37.10 | jacques | well that was fun |
| 10:37.25 | NAiL | dyoung: Very unlikely. |
| 10:40.15 | dyoung | http://groups.yahoo.com/group/nslu2-linux/message/2452 |
| 10:40.29 | dyoung | I dont have the issue, but rwhitby does. |
| 10:40.51 | dyoung | basically you need to add a 100k or so resistor between pin 1 and 2 of the slug serial port. |
| 10:41.54 | NAiL | hmm... Fair enough. Then I'll just have to find a resistor |
| 10:42.12 | dyoung | it escapes me why it works for me and not others. |
| 10:42.26 | dyoung | I guess its possible my cable has one already. |
| 10:42.29 | jacques | I didn't need any resistors |
| 10:43.57 | dyoung | jacques, nod |
| 10:44.19 | dyoung | NAiL, http://nslu2-linux.thestuffguy.com/gallery/Serial_adapters |
| 10:44.27 | NAiL | Hmm.. I've got an alternative cable, but I don't know which leads are what... |
| 10:44.31 | dyoung | I use the RS 170-0787 |
| 10:44.34 | NAiL | dyoung: I've seen it. |
| 10:44.56 | dyoung | nail, what sort of wires do you have coming out of your mystery calbe? |
| 10:44.58 | dyoung | cable? |
| 10:45.20 | NAiL | what sort of? There's one blue, one white and one black ;) |
| 10:45.28 | dyoung | if you look at my serial adapters gallery, you'll see some black things that are marked 152, those are resistors. |
| 10:45.45 | dyoung | so its totally possible that this cable is pulled up already. |
| 10:45.58 | dyoung | Do you have a Voltmeter handy? |
| 10:46.04 | NAiL | infact I do. |
| 10:46.16 | NAiL | or.. "handy" is an overstatement. |
| 10:46.19 | NAiL | I'll dig it up |
| 10:46.25 | dyoung | is it a PL2303 device? |
| 10:47.04 | NAiL | yes |
| 10:47.19 | dyoung | Okay, this is the usual way to figure it out. |
| 10:47.35 | dyoung | figure out which of your 3 wires is ground. |
| 10:47.41 | dyoung | we'll call that wire A. |
| 10:48.02 | dyoung | then measure the voltage between A and B then A andn C. |
| 10:48.11 | dyoung | one will be 0V one will be 3.3V |
| 10:48.40 | dyoung | the one thats 3.3V is the TX line that will go to the Slug RX Line (pin2) |
| 10:49.09 | dyoung | the other one is the RX which will go to the slug tx line (pin3) |
| 10:50.28 | NAiL | ok, between white and black I get 3,3. But which do I figure out which one is 3.3? I don't know which end is what of this voltmeter ;) |
| 10:50.45 | dyoung | you need to figure out which one is ground. |
| 10:51.01 | dyoung | hmm |
| 10:51.12 | dyoung | you can usually look at the board and figure it out |
| 10:51.34 | NAiL | hehe, yeah. |
| 10:51.46 | dyoung | sorry, I cant think of a easy shortcut for that. |
| 10:51.56 | NAiL | no, I'll try to figure it out. |
| 10:52.11 | NAiL | There's three annoying leds on the board that makes it hard though |
| 10:52.34 | NAiL | whoever thought of adding a ultra-brite (tm) blue led on this board should have been shot. Thrice. |
| 10:52.58 | dyoung | yo could put your voltmeter into resistance mode and measure against the common side of your LED's to the black and white. |
| 10:54.04 | dyoung | Or, the negative probe will be the ground side when you measure volts accross the white/black. Maybe. |
| 10:56.57 | ep1220 | [g2]: I am back |
| 10:57.08 | [g2] | ep1220, wb |
| 10:57.17 | NAiL | hmm.. |
| 10:57.25 | [g2] | I hope your lunch was enjoyable |
| 10:57.29 | [g2] | and delicious |
| 10:57.41 | NAiL | white/black, 3.3v. Blue/Black, 2.8v? |
| 10:58.02 | ep1220 | [g2]: it was OK, nothing special |
| 10:58.26 | dyoung | white blue? |
| 10:58.38 | ep1220 | [g2]: I tried to add code so the GPIO14/15 output the 33Mhz CLK |
| 10:58.43 | [g2] | well if we have special everyday then it becomes ordinary right ? :) |
| 10:58.44 | ep1220 | but did not work |
| 10:59.07 | NAiL | white/blue 0.0-something |
| 11:00.08 | NAiL | Fun. The blue led goes crazy when I touch the black or white lead :) |
| 11:00.33 | dyoung | those were the only wires coming out of the thing? |
| 11:01.41 | NAiL | yeah |
| 11:03.27 | *** join/#openjtag ChanServ (ChanServ@services.) |
| 11:03.27 | *** join/#openjtag jbowler-zzz (~jbowler@jbowler.nslu2-linux) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag pjames (~jcouzens@mail.lost-carrier.com) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag Tiersten (tman@Tiersten.nslu2-linux) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag ep1220 (~NN@gate.epatec.at) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag nslu2-log (dyoung@netblock-66-159-209-60.dslextreme.com) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag rwhitby (~rwhitby@CPE-60-226-247-226.sa.bigpond.net.au) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag NAiL (~repvik@banan.dataparty-mn.biz) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag Jacmet (~jacmet@50C59F0E.flatrate.dk) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag VoodooZ_Log (~VoodooZ@CPE00c0f021712c-CM014110003785.cpe.net.cable.rogers.com) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag dyoung (~dyoung@cpe-24-94-88-62.hawaii.res.rr.com) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag ka6sox-office (~tking@netblock-66-159-209-52.dslextreme.com) [NETSPLIT VICTIM] |
| 11:03.27 | *** join/#openjtag jacques (~jacques@jacques.nslu2-linux) [NETSPLIT VICTIM] |
| 11:03.27 | *** mode/#openjtag [+o ChanServ] by irc.freenode.net |
| 11:03.59 | dyoung | Blue is probably RX |
| 11:04.00 | dyoung | but I take no responsibility for blown up calbes, etc |
| 11:04.00 | dyoung | :-) |
| 11:04.29 | NAiL | Isn't there a pinout of the prolific anywhere? I see the blue goes directly to a pin there |
| 11:04.44 | dyoung | just moment |
| 11:04.45 | NAiL | (through 17 gazillion vias) |
| 11:04.54 | dyoung | what pin number? |
| 11:06.33 | NAiL | Kinda hard to say, the top of the chip doesn't have anything on it... |
| 11:06.48 | dyoung | no dimple or.. ? |
| 11:07.47 | dyoung | is it the 5th pin from something? |
| 11:08.10 | dyoung | from one of the ends? |
| 11:09.24 | [g2] | ep1220, did you assemble that code bewoolie sent or just use the hex ? |
| 11:09.34 | ep1220 | [g2]: OK, when I just enable GPIO14/15 and donot try to setup the clock I get the 33Mhz |
| 11:09.52 | ep1220 | g2: the code bewoolie mailed me did not work |
| 11:10.29 | ep1220 | So i assembled my own. |
| 11:11.16 | ep1220 | based on our 1st prog |
| 11:11.58 | ep1220 | For experiments it is best to toggle the green status LED. |
| 11:12.07 | ep1220 | This signal can be seen on R53 |
| 11:14.49 | ep1220 | [g2]: Just mailed it to You |
| 11:14.54 | [g2] | ok |
| 11:15.40 | ep1220 | I just do not know how to interpret the result :-( |
| 11:16.03 | [g2] | so the code you just sent toggles the LED line at 5Mhz |
| 11:16.39 | [g2] | ah.. and you're running it from RedBoot right ? |
| 11:16.46 | ep1220 | Yes. measured between 2 rising edges of GPIO1 |
| 11:16.56 | ep1220 | No, directly from RESET |
| 11:17.13 | ep1220 | No bootloader running before |
| 11:17.30 | dyoung | thats why ep1220 is my idol |
| 11:17.33 | [g2] | I'd try cache on in Redboot and run it from redboot |
| 11:17.42 | [g2] | I'll bet the cache is off :) |
| 11:17.56 | ep1220 | I load it into the cache |
| 11:18.19 | dyoung | [g2] make joke. :-) |
| 11:18.45 | dyoung | NAiL, I'm running out of battery. |
| 11:19.07 | dyoung | I have the ds open now if you need any info. |
| 11:19.16 | [g2] | ep1220, are you running from the main icache or mini icache |
| 11:19.32 | ep1220 | mini, maybe the minicache is not 0 WS (??) |
| 11:20.03 | [g2] | can you try just load and running the program from Redboot ? |
| 11:20.28 | NAiL | Note to self, do *not* attempt to probe a USB-device attached to Window |
| 11:20.30 | NAiL | s |
| 11:20.47 | [g2] | maybe you don't have the tftp server setup |
| 11:20.48 | ep1220 | g2: So far I only have an assembly listing. |
| 11:21.01 | NAiL | dyoung: Yes, it has a dimple |
| 11:21.06 | ep1220 | I have no tftp setup |
| 11:21.15 | ep1220 | but I can serial into redboot |
| 11:21.22 | dyoung | nail, the dimple is pin1 |
| 11:21.33 | dyoung | its on the left side |
| 11:21.51 | dyoung | you count down from tht pin, to pin 14, then accross to the lower right which is pin 15. |
| 11:21.59 | dyoung | the one accross from the dimple is pin 28. |
| 11:21.59 | [g2] | dyoung, is ymodem supported in the nslu2 Redboot ? |
| 11:22.12 | dyoung | Funny you should ask that. |
| 11:22.26 | dyoung | the -m supports xmodem ymodem tftp http |
| 11:22.28 | dyoung | only. |
| 11:22.32 | dyoung | no zmodem. |
| 11:23.15 | [g2] | Ok.. that's cool... ep1220 can x or y modem a tiny program to redboot |
| 11:23.46 | NAiL | the blue one is connected to the chip through a via, so it's kinda hard to seee. But... I buzzed from black to the chip. Everything buzzes. That'd be gnd, right? |
| 11:24.24 | [g2] | ep1220, you've got the object code genterated from that program right ? |
| 11:24.32 | dyoung | the chip has a few grounds on it, pin 7 and 21 are ground. |
| 11:24.58 | dyoung | pin 18 is the 3V3 gnd |
| 11:25.14 | NAiL | how many pins on each side? |
| 11:25.40 | NAiL | I'm farsighted, which doesn't help atm |
| 11:26.08 | dyoung | 14 pins on each side. |
| 11:26.48 | [g2] | ep1220, if you just hexdump/hexedit the object code you can see where the first instruction offset starts |
| 11:26.50 | ep1220 | g2: No. Just a listing |
| 11:27.14 | [g2] | what command line are you using with gas ? |
| 11:27.21 | ep1220 | the code is fully relocatable. |
| 11:27.25 | ep1220 | I did not use gas |
| 11:27.48 | [g2] | nodd |
| 11:28.03 | dyoung | ... I have gas..... |
| 11:28.09 | [g2] | :) |
| 11:28.27 | [g2] | the 2 things I'd try are: |
| 11:28.37 | [g2] | 1) run from redboot with the cache one |
| 11:29.06 | dyoung | nail, shutdown sequence initiated. |
| 11:29.14 | [g2] | 2) unroll the loop and put like 10 toggles before the branch |
| 11:29.23 | ep1220 | OK, I have a .o now |
| 11:29.31 | NAiL | dyoung: Going to bed? |
| 11:29.36 | dyoung | Critical hit: heat sink. |
| 11:29.51 | dyoung | Internal Ammunition explosion detected. |
| 11:29.54 | dyoung | http://tech.prolific.com.tw/visitor/fcabdl.asp?fid=33962507 |
| 11:29.59 | [g2] | 3) insert a NOPs or ORR rX, rx,rx instructions as place holders |
| 11:31.03 | dyoung | nail, soon. I have around 5minutes of battery left. |
| 11:31.21 | [g2] | one day ends... another begins ..... |
| 11:31.22 | NAiL | dyoung: Ok, I'll try to figure out the rest. Thanks :) |
| 11:31.23 | ep1220 | g2: the assembler can output aout. Is this OK for redboot |
| 11:31.51 | [g2] | ep1220, it's just binary, you'll need to see the offset of the code though |
| 11:34.43 | dyoung | NAiL, If you remember that TX is asserted idle mode you shouldbe able to figure it out |
| 11:38.06 | ep1220 | g2: Offset of the code == startaddress ? |
| 11:39.24 | ep1220 | g2: Just now I will add a nop (mov r2,r2) into the loop and check how it affects timing |
| 11:40.03 | [g2] | it's the start address you want |
| 11:40.41 | ep1220 | I can run the code anywere, there are no absolute calls/jumps in there |
| 11:40.52 | [g2] | nod |
| 11:41.14 | *** join/#openjtag prpplague (~billybob@216.110.99.130) |
| 11:41.37 | [g2] | I'd do something like |
| 11:42.08 | [g2] | load -r -v -b 0x1000000 -m YMODEM |
| 11:42.19 | [g2] | send the code |
| 11:42.43 | [g2] | exec 0x1000034 or whatever the first instruction address is |
| 11:42.54 | [g2] | or go 0x1000034 |
| 11:44.39 | ep1220 | g2: How do I get into redboot. I see a '+' and a little later the startup messages |
| 11:44.52 | [g2] | just hit the ctrl-c |
| 11:45.02 | [g2] | after the '+' |
| 11:56.14 | [g2] | ep1220, luck with Redboot ? |
| 12:00.01 | ep1220 | g2: No, I see what it sends, but it ignores what I type |
| 12:01.32 | [g2] | Have you used serail for input before ? |
| 12:02.16 | ep1220 | Yes. I also have already been in redboot (just didnot remember the key) |
| 12:03.11 | [g2] | you can also get into Redboot via Telnet |
| 12:03.40 | [g2] | it's a tighter window, but it sounds like the receive side of your serial isn't working |
| 12:03.55 | [g2] | or your terminal emulator isn't properly configured |
| 12:04.11 | [g2] | like it's setup for flow-control |
| 12:12.43 | ep1220 | I get me a cup of tea and then try again :-) |
| 12:18.03 | NAiL | heh, I think I figured out which cable is rx... If I touch the blue wire, I get lots of garbage :-P |
| 13:00.31 | VoodooZ_Work | Good work guys. I'm keeping my fingers crossed we can figure out a way to get a "free" speed boost. |
| 13:01.48 | NAiL | I wouldn't mind a free speed boost either. But I think it's unlikely ;) |
| 13:03.21 | VoodooZ_Work | yeah, I'm just a dreamer :) |
| 13:04.53 | NAiL | Nah, A Visionary. Sounds better :D |
| 13:05.24 | VoodooZ_Work | ok. if you insist. :) |
| 13:33.31 | *** join/#openjtag Tiersten_ (tman@Tiersten.nslu2-linux) |
| 14:20.06 | ep1220 | g2: ping |
| 14:24.01 | VoodooZ_Work | ep1220: Found anything more? |
| 14:24.15 | ep1220 | I had a broken wire |
| 14:24.20 | ep1220 | now I am in redboot |
| 14:24.25 | ep1220 | loaded the code |
| 14:24.39 | ep1220 | when i exec it does not wha it should :-( |
| 14:24.53 | ep1220 | i use -b 10000000 |
| 14:24.56 | VoodooZ_Work | odd. |
| 14:25.02 | [g2] | ep1220, poing |
| 14:25.12 | VoodooZ_Work | [g2] to the rescue! |
| 14:25.13 | ep1220 | but it says it loads to 989680 |
| 14:25.28 | ep1220 | when i dump: x -b 10000000 |
| 14:25.41 | ep1220 | the output starts at 989680 |
| 14:25.46 | ep1220 | is this normal ? |
| 14:26.08 | [g2] | I think you have too many zeros :) |
| 14:26.31 | [g2] | reset |
| 14:26.45 | ep1220 | done; seesm I have 1 too much |
| 14:27.00 | VoodooZ_Work | been there! |
| 14:27.07 | [g2] | yes |
| 14:27.28 | [g2] | it's 10 5-zeros |
| 14:27.36 | [g2] | which is 16 MB |
| 14:27.52 | [g2] | 5-zeros being 2**20 |
| 14:31.34 | [g2] | ep1220, is that helping ? |
| 14:32.19 | ep1220 | Well, now the addresses match, but still no blinking LED |
| 14:32.39 | ep1220 | When I dump, I see the LSB first, is this OK ? |
| 14:33.23 | [g2] | I think you should be seeing MSB |
| 14:33.42 | [g2] | are you go -b 0x1000000 ? |
| 14:33.54 | [g2] | it should probably be -b 0x100034 |
| 14:34.04 | [g2] | objdump is your friend |
| 14:34.18 | ep1220 | no I do exec 0x100000 |
| 14:34.30 | [g2] | that's why it's not working |
| 14:34.56 | ep1220 | the first codeword is at 0x100000 |
| 14:35.15 | [g2] | then it should work |
| 14:35.37 | [g2] | you should see the same instructions in memory as the hex |
| 14:37.50 | ep1220 | Is identical |
| 14:37.55 | ep1220 | now I use go |
| 14:38.30 | ep1220 | this gives: T050f:00100008;0d:01ff2fec;#e3 |
| 14:39.37 | [g2] | I think that's the debugger |
| 14:40.02 | [g2] | your not executing stuff right or you compiled it in LE mode |
| 14:46.45 | ep1220 | yes it is LE |
| 14:47.13 | ep1220 | I have to make a phone-call now and will try again afterwards |
| 14:48.11 | [g2] | nod. |
| 14:51.31 | jacques | did you guys see what jbowler-zzz said? |
| 14:51.49 | jacques | <jbowler-away> Am I missing something? That code seems to be writing to the physical address of the GPIO within the loop, but writing the GPIO is very slow - it won't show the core clock speed at all. |
| 14:51.49 | jacques | <jbowler-away> mov r0, #0 |
| 14:51.49 | jacques | <jbowler-away> 0: subs r0, r0, #1 |
| 14:51.49 | jacques | <jbowler-away> bne 0b |
| 14:51.49 | jacques | <jbowler-away> Should take about 32s with a 266MHz core clock (2^33 instructions) |
| 14:53.06 | [g2] | jacques, I saw it and I didn't see any reference in the Developers Manual about the timing for writing to that register |
| 14:55.23 | jacques | `k just wanted to make sure you saw it |
| 14:55.36 | ep1220 | g2: big endian works better :-) |
| 14:56.31 | [g2] | ep1220, :) |
| 14:56.32 | ep1220 | is still slower, here I see 4.12MHz |
| 14:57.58 | [g2] | is cache on ? |
| 14:58.34 | ep1220 | How can I check ? |
| 14:59.08 | ep1220 | made a mistake: it is 6.58Mhz |
| 15:00.02 | [g2] | cache on |
| 15:01.40 | [g2] | 6.58 is better than 4.12 :) |
| 15:03.50 | ep1220 | cache on gives "illegal command jhddcache" |
| 15:06.43 | ep1220 | Using "cache ON": I get 7.25MHz |
| 15:13.06 | ep1220 | I now added a move r3,r2 to the loop |
| 15:13.16 | ep1220 | gives 5,45MHz |
| 15:13.59 | ep1220 | if i calc the delta and asume this is one cycle i get a coreclock of approx- 44Mhz |
| 15:14.00 | ep1220 | As jacques said above: I start to believe that writing to GPIO is not a valid way to find out the internal clockrate |
| 15:15.12 | ep1220 | Maybe a larger count-down inside the loop is better |
| 15:15.57 | [g2] | we have a loop inside |
| 15:16.18 | [g2] | a million 0x100000 |
| 15:17.11 | [g2] | that'll give an approximate freq in MIPS (if we use 0x80000 |
| 15:26.59 | ep1220 | Added a loop |
| 15:27.09 | ep1220 | count down 4096 to 0 |
| 15:27.29 | ep1220 | I assume the sub and branch take 1 cycle each |
| 15:27.58 | ep1220 | Now the GPIO is low/high for 62microseconds (each phase) |
| 15:28.20 | ep1220 | This gives a clock rate of 132,12MHz |
| 15:35.51 | VoodooZ_Work | THe question is: how can we modify this? :) |
| 15:38.13 | ep1220 | Is the assumptin: branch takes one cycle correct ? |
| 15:38.24 | ep1220 | It it were 3 cycles, we were at 266MHz |
| 15:39.55 | VoodooZ_Work | [g2] would be able to answer that. |
| 15:40.14 | VoodooZ_Work | I'm more of a 8 bit micro guy. |
| 15:41.59 | ep1220 | There is a register named EXP_CNFG0 |
| 15:42.31 | ep1220 | I ahve an app-note which says that bits 23:21 in there can be used to under-clock a part. |
| 15:43.04 | VoodooZ_Work | interesting! |
| 15:43.16 | ep1220 | It also allows to select the PCI clock |
| 15:43.21 | ep1220 | 33 or 66 |
| 15:43.21 | VoodooZ_Work | We'd have to figure out what it's being set at in the slug. |
| 15:43.27 | VoodooZ_Work | nice. |
| 15:44.12 | ep1220 | It is at 0xC4000020 |
| 15:45.20 | VoodooZ_Work | I'd have to browse the different data sheet as I'm a total ARM noob |
| 15:46.51 | ep1220 | tried to dump it in Redboot -> gives a strange line, maybe an exception (?) |
| 15:48.07 | ep1220 | I'll be back later |
| 15:48.25 | VoodooZ_Work | ok, we'll have to share your findings with the others. |
| 15:51.07 | [g2] | APEX can dump it :) |
| 15:51.52 | VoodooZ_Work | hehe |
| 15:52.19 | VoodooZ_Work | A link on the performance wiki page would be useful |
| 15:52.41 | VoodooZ_Work | Did you see ep1220's assumption above? Can you confirm? |
| 15:53.02 | VoodooZ_Work | (branches take 1 cycle or more?) |
| 15:53.27 | [g2] | I'd have to check |
| 16:32.27 | *** join/#openjtag JMunakra (~JMunakra@dsl-084-059-203-219.arcor-ip.net) |
| 16:43.39 | *** join/#openjtag ByronT_ (ByronT@ByronT.nslu2-linux) |
| 18:10.51 | *** join/#openjtag nslu2-log_ (dyoung@dyoung.nslu2-linux) |
| 18:34.56 | *** join/#openjtag ByronT (ByronT@ByronT.nslu2-linux) |
| 18:40.56 | ep1220 | [g2]-away: the appNote is IXP42X Boot-up Options; Intel doc number from Filename: 25406701; printed inside: 254067-002 |
| 18:52.42 | VoodooZ_Lunch | interesting read. |
| 18:53.42 | VoodooZ_Lunch | Makes me wonder if configuring the 266 part as if it was a 533 part but configured for 266 operation would give us 133Mhz? |
| 18:55.00 | VoodooZ_Lunch | [g2]-away: if you get a chance can you dump the content of that config reg in Apex? |
| 19:00.36 | jacques | <PROTECTED> |
| 19:00.45 | jacques | interesting indeed |
| 19:02.25 | VoodooZ_Lunch | yeah, I doubt it's that simple but I'm curious to see what our beloved slug is configured. |
| 19:02.52 | jacques | I just want to get to the bottom of this one way or another |
| 19:02.58 | VoodooZ_Lunch | yeah, me too. |
| 19:03.11 | jacques | and if we end up being able to increase our performance in the process, I will be very heppy |
| 19:03.14 | jacques | happy |
| 19:03.17 | VoodooZ_Lunch | the DS says these config bits can be configured with weak pullups. |
| 19:03.24 | VoodooZ_Lunch | damn right! |
| 19:03.35 | VoodooZ_Lunch | that would help my computer vision thread quite a bit! |
| 19:03.42 | jacques | I bet it would |
| 19:05.26 | jacques | hmm, they still don't show a way to actually underclock a 266MHz part |
| 19:07.21 | VoodooZ_Lunch | yeah, but sometimes they don't show all "possible" combinations in those tables. I'd love to try a few other values. |
| 19:07.50 | jacques | VoodooZ_Lunch, agreed |
| 19:08.07 | jacques | first we ned to know what it's set to |
| 19:08.12 | VoodooZ_Work | oops. forgot to change nick. |
| 19:08.18 | VoodooZ_Work | yep. |
| 19:08.28 | jacques | and if it's not 1,0,0 I will be very curious |
| 19:08.33 | VoodooZ_Work | [g2] said he could read that value using Apex. |
| 19:08.38 | VoodooZ_Work | indeed! |
| 19:08.40 | jacques | if it's 0,1,1 I will think we've found the answer |
| 19:08.41 | VoodooZ_Work | I'm all excited! |
| 19:09.02 | VoodooZ_Work | Nothing like a good overclock to make me smile! |
| 19:10.16 | VoodooZ_Work | unless it's possible to read it using redboot? ep1220 said he had errors. |
| 19:12.06 | VoodooZ_Work | a simple 10k pull-down should do it then. |
| 19:12.20 | VoodooZ_Work | Unless we don't have easy access to those pins? |
| 19:12.31 | NAiL | That'd be a drag |
| 19:12.37 | VoodooZ_Work | damn right! |
| 19:18.42 | NAiL | But I'd imagine it would generate a lot of heat though |
| 19:18.42 | NAiL | hmm... ping me someone? |
| 19:19.59 | VoodooZ_Work | yeah, but I would still like to do it as my robot doesn't run for that long anyways. |
| 19:20.26 | VoodooZ_Work | I tried doing a dump of that memory area from redboot but it keeps giving me weird stuff. |
| 19:20.38 | VoodooZ_Work | someone know how to stop the dump command? |
| 19:21.19 | NAiL | BE/LE? |
| 19:23.31 | VoodooZ_Work | well, the default slug redboot mode. |
| 20:27.32 | [g2] | dump 0xc4000020 |
| 20:27.32 | [g2] | c4000020: 00 41 ff ee 00 00 00 00 00 00 09 fc 00 00 00 00 .A??.... ...?.... |
| 20:27.32 | [g2] | c4000030: 00 00 29 46 08 04 40 c4 00 00 00 00 00 00 00 00 ..)F..@? ........ |
| 20:27.32 | [g2] | c4000040: 10 00 00 00 96 54 22 21 ff ff ff ec ff ff ff ff .....T"! ???????? |
| 20:35.24 | [g2] | jacques, I think that 4 makes it 0, 1 0 |
| 20:41.12 | ep1220 | unfortunately the bits setting the clockfrequency can not be written to |
| 20:43.19 | [g2] | it's the pull ups right ? |
| 20:43.27 | ep1220 | nod |
| 20:43.54 | ep1220 | I already tried 2 values, no change in execution speed. |
| 20:44.53 | [g2] | tried writing two values ? |
| 20:45.16 | ep1220 | yes. 1,0,0 and 0,1,1 |
| 20:45.26 | [g2] | does the write work ? |
| 20:45.52 | ep1220 | it does not crash. |
| 20:46.06 | ep1220 | But i did a "blind" write. |
| 20:46.07 | [g2] | I meant did you read back the value ? |
| 20:46.14 | ep1220 | no |
| 20:46.48 | ep1220 | just added code to set these bits before the toggling |
| 20:47.24 | ep1220 | does the dump command also work in redboot ? |
| 20:47.40 | ep1220 | I tried x -b 0xc4000020 -l 4 |
| 20:47.47 | ep1220 | gave strange repsonse |
| 20:48.14 | *** join/#openjtag beewoolie-afk (~beewoolie@florence.buici.com) |
| 20:48.57 | [g2] | beewoolie-afk, in da house ! |
| 20:49.36 | ep1220 | g2: did You verify how this reg is set on your faster machine (avila ?) |
| 20:50.09 | [g2] | ep1220, no, not yet |
| 20:50.21 | [g2] | I forget which boot loader I've got there (probably APEX) |
| 20:51.12 | [g2] | I was going to try changing the value on the non-fat slug with APEX before the boot |
| 20:53.39 | [g2] | beewoolie-afk, is there a write other than fill ? |
| 20:56.23 | beewoolie-afk | [g2]: yo man. did you follow the thread with ep1220? |
| 20:56.47 | beewoolie-afk | [g2]: What are you looking to do? |
| 20:56.55 | beewoolie-afk | [g2]: re fill tha is |
| 20:56.57 | [g2] | beewoolie-afk, I think so, I didn't do the math on the instructions |
| 20:57.07 | [g2] | can you fill more than a byte ? |
| 20:57.44 | [g2] | fill 0x0080ffee 0xC4000020+4 |
| 20:58.01 | [g2] | fill 0x0081ffee 0xC4000020+4 |
| 20:59.34 | *** join/#openjtag JMunakra (~JMunakra@dsl-084-059-203-219.arcor-ip.net) |
| 21:00.06 | beewoolie-afk | [g2]: That should work. |
| 21:00.30 | beewoolie-afk | Oh, you are wondering if the source can be more than one byte. I don't recall. I think that I never made that work. |
| 21:00.56 | [g2] | nod. it didn't :) |
| 21:01.51 | [g2] | I'd think jacques would have mmap'ed that reg by now |
| 21:02.34 | dyoung-zzzz | We've talked about this appnote and register before. |
| 21:03.51 | dyoung-zzzz | what we really need to do is see if CFG1 and/or CFG0 are pulled up. |
| 21:03.55 | dyoung-zzzz | externally. |
| 21:04.22 | [g2] | I've been wanting to dump those initial values |
| 21:04.34 | [g2] | actuall all the values from the pull ups |
| 21:04.44 | [g2] | it's in CPSR15 iirc |
| 21:04.57 | [g2] | that's one of the first values read |
| 21:07.13 | [g2] | hey |
| 21:07.25 | [g2] | dump 0xc4000020 |
| 21:07.25 | [g2] | c4000020: 00 81 ff ee |
| 21:07.27 | [g2] | BINGO |
| 21:07.44 | [g2] | ep1220, did you see that ? |
| 21:07.52 | ep1220 | yes |
| 21:08.39 | ep1220 | now you can run your benchmark |
| 21:08.56 | [g2] | ep1220, that's the *avila* board :) |
| 21:09.21 | ep1220 | :-( I thought You could modify them on the slug |
| 21:09.56 | ep1220 | looks like we are back on searching traces |
| 21:10.17 | ep1220 | the Flash includes only A21 |
| 21:10.48 | [g2] | it's on the expansion address lines |
| 21:13.49 | ep1220 | right, just found A23 in the wiki, So r64 pulls-down A23 |
| 21:16.33 | [g2] | ep1220, which wiki page ? |
| 21:16.43 | ep1220 | http://www.nslu2-linux.org/wiki/Info/FlashExpansion |
| 21:16.48 | [g2] | thx |
| 21:18.04 | [g2] | ep1220, I thought it just read the lower 16 for the ixp420 |
| 21:19.41 | ep1220 | as I understand this appnote: at reset it reads A0:30 into the reg at c4000020 |
| 21:21.26 | ep1220 | a21:23 set the clock frequency |
| 21:21.54 | [g2] | I'm probably remembering wrong |
| 21:21.55 | dyoung-zzzz | Okay I remember now. |
| 21:22.10 | dyoung-zzzz | page 326 of 25248005.pdf |
| 21:23.27 | [g2]-dinner | so addr 23 and 22 need tweaked |
| 21:23.44 | [g2]-dinner | and the 010 is undoco |
| 21:23.54 | dyoung-zzzz | Well, I dunno, according to that table the 266Mhz can only be 266, |
| 21:24.00 | ep1220 | i verified R64 is a pull-down, but it might be connected to the FLASH only |
| 21:24.26 | dyoung-zzzz | but 101 110 and 111 are not documented. |
| 21:24.41 | dyoung-zzzz | plus what [g2] said/ |
| 21:25.09 | beewoolie-afk | My read of those registers is that these are just the settings for memory timing. Moreover, this register is just for the flash rom. |
| 21:25.49 | beewoolie-afk | Ooops. Wrong page. |
| 21:25.58 | dyoung-zzzz | beewoolie-afk, the table is titled, "Intel XScale Core Speed Expansion Bus Configuration Strappings" |
| 21:26.46 | beewoolie-afk | But there isnt anything in the manual about 133MHz. |
| 21:27.04 | dyoung-zzzz | exactly my point. |
| 21:27.21 | beewoolie-afk | Yet, that doesn't mean that they don't detune parts. |
| 21:27.21 | ep1220 | i think the table does not necessarily imply that other settings do not work |
| 21:27.27 | dyoung-zzzz | I looked at this (in the manual only) before and thought tht 133Mhz was not supported. |
| 21:27.47 | dyoung-zzzz | ep1220, Right. and I'm happy that someone is looking at it physically! |
| 21:27.54 | beewoolie-afk | Well, we could try psuing the value to 0b100 to see if it makes the CPU run faster. |
| 21:28.24 | ep1220 | the APPnote says these bits can not be written to |
| 21:28.47 | ep1220 | (i already tried, was not faster, but not 100% sure my prog was correct) |
| 21:28.52 | dyoung-zzzz | if we compare against the 533Mhz part, it says 011 is the half speed; but 011 for the 266 is still 266. |
| 21:28.55 | beewoolie-afk | Hmm. And we're already using 0b100 |
| 21:29.20 | ep1220 | g2: read out 0b010 |
| 21:29.36 | ep1220 | this is nowhere in the table |
| 21:30.03 | beewoolie-afk | Didn't you read out 0x0081ffee? That looks like 0b100 to me. |
| 21:30.19 | ep1220 | this was on the avila |
| 21:30.25 | beewoolie-afk | Ah. |
| 21:30.39 | beewoolie-afk | so, perhaps we could change one of the resistors? |
| 21:30.39 | ep1220 | it was 0x0041ffee on the slug |
| 21:30.56 | beewoolie-afk | Very interesting. |
| 21:30.58 | ep1220 | when we find them: yes :-) |
| 21:31.06 | ep1220 | but need change 2 to get 100 |
| 21:32.40 | beewoolie-afk | Or make it zero. |
| 21:33.45 | beewoolie-afk | Making it zero would probably be easier. |
| 21:33.58 | ep1220 | zero is undocumented as well |
| 21:34.25 | ep1220 | based on the table I'd say unless a23 is 1 we might not get full speed |
| 21:34.57 | beewoolie-afk | Perhaps. All zeros is full speed in every other case. |
| 21:34.59 | dyoung-zzzz | ep1220, look at that page 326, it documents 100 000 001 011 for being 266Mhz. |
| 21:34.59 | ep1220 | in the appnote a23 is named cfg_en_n |
| 21:35.45 | dyoung-zzzz | (thats 23:21) |
| 21:37.24 | ep1220 | Funny, the Names of these bits are different than in the appNote |
| 21:38.03 | ep1220 | (on page 325) |
| 21:38.59 | ep1220 | the devmanual is the more recent document |
| 21:39.09 | beewoolie-afk | According to the pin schedule, these are (23-21) J24 G26 F26 |
| 21:41.39 | *** join/#openjtag VoodooZ (~VoodooZ@CPE00c0f021712c-CM014110003785.cpe.net.cable.rogers.com) |
| 21:44.06 | beewoolie-afk | I'm not so good at the topological twistings here. |
| 21:46.22 | dyoung-zzzz | tracking traces is painful |
| 21:46.33 | ep1220 | It is close to midnight here, so i too am not in the mood for tracking traces :-) |
| 21:46.51 | ep1220 | I give it a try tomorrow. |
| 21:46.57 | VoodooZ | thanks ep1220 |
| 21:47.04 | ep1220 | g'night all |
| 21:47.05 | dyoung-zzzz | if I'm feeling brave, I'll look at it late |
| 21:47.06 | dyoung-zzzz | r |
| 21:47.19 | dyoung-zzzz | printit out gigantic. |
| 21:47.39 | VoodooZ | cool! |
| 21:47.53 | beewoolie-afk | It looks to me to be a via to the other side of the board. |
| 21:48.24 | *** part/#openjtag ep1220 (~NN@gate.epatec.at) |
| 21:49.20 | beewoolie-afk | It would be a cinch if I had one of those bar boards. |
| 21:49.24 | beewoolie-afk | s/bar/bare/ |
| 21:49.32 | beewoolie-afk | A DVM would make quick work of the search. |
| 21:50.45 | beewoolie-afk | VoodooZ: if the zeros are pull downs and the 1's are floating, then it should be simple to bring up the high bit. |
| 21:51.26 | VoodooZ | yeah. |
| 21:52.17 | VoodooZ | the DS says to use 10K weak pull-downs to set the bits. |
| 21:52.33 | VoodooZ | as the pins are internall pulled-up IIRC |
| 21:56.44 | dyoung-zzzz | ok I printed it it super gigantic. |
| 21:56.54 | beewoolie-afk | ginormous? |
| 21:57.02 | dyoung-zzzz | Gimangous! |
| 21:57.25 | beewoolie-afk | BTW, if this works, it will make the slug a *rockin* platform. |
| 21:57.46 | dyoung-zzzz | I guess I need to either take a slug with me, or print out the populated board too |
| 21:57.49 | beewoolie-afk | I'll have to buy a few extras in case it makes linksys mad. |
| 21:58.22 | NAiL | dyoung-zzzz: Going to read till you fall asleep? ;) |
| 22:00.53 | [g2] | hey can you guys join me in #nslu2 ? |
| 22:56.15 | *** join/#openjtag basstech (~basstech@ip68-111-62-90.br.no.cox.net) |
| 23:43.48 | *** join/#openjtag Tiersten (tman@Tiersten.nslu2-linux) |