00:00:26 right 00:01:05 Yup. The tree widget stuff itself is about 70k of tickle code. 00:20:15 rayh: can you separate that tcl code? 00:20:29 so we can distribute it to people who lack it? 00:23:10 I believe I can separate it out. 00:24:04 I'll try that tomorrow. 00:27:36 hmmm - have you read the email from JimJames on the emc-users list? 00:29:09 I did and there is NOT a machine on/off pin. 00:29:26 must be a different email ;) 00:29:57 there's one that just came in a couple of minutes ago, saying that the stg config wasn't a choice for him 00:30:05 the closest you come in emc2 is axis.0.amp-enable-out 00:30:16 or any of the other axes that are defined. 00:30:59 hmm.. JimJames needs to wait a few days 00:31:05 indeed 00:31:06 I'll make a basic ini 00:31:24 is the configs/stg dir just a placeholder, or can it actually work? 00:31:44 it holds the hal files so far 00:31:48 no ini yet 00:31:52 ah - ok 00:31:58 so that might be a bit much for a newbie ;) 00:32:03 heh 00:32:13 the README is oh so helpful as well ;) 00:32:36 fix it :P 00:32:43 stop complaining about it :) 00:32:46 as is the Windows information when you mouse over it "Type: file" 00:33:18 I'm not the person to fix it - I have no STG card or experience :P 00:34:02 I loaned my stg out a while back. 00:34:32 Does the standard stg pinouts include an amp enable for each axis? 00:34:50 if so that is the pin to refer him to. 00:35:42 rayh: the stg_io.hal and stg_motion.hal are done in such a way to emulate emc1 behaviour fully 00:35:57 but his problem is that there isn't an ini, and for a newbie that's a bit much 00:36:12 Oh. Okay. 00:36:28 I'll do it tomorrow (lateron today :) 00:36:38 right - cahnging MOTION=motmod to MOTION=-hal_stg didn't work ;) 00:36:41 changing 00:36:58 rayh has quit 00:37:26 well - time to go out to a party. wee you later 00:37:31 see, that is :) 00:37:45 SWPadnos is now known as SWP_Away 00:40:42 jmk_dinner is now known as jmkasunich 06:33:49 SWPadnos has joined #emc-devel 06:34:58 SWP_Away has quit 06:34:58 LawrenceG has quit 06:35:56 SWP_Away has joined #emc-devel 06:35:56 LawrenceG has joined #emc-devel 06:39:11 [Global Notice] Hi all. We just experienced routing problems with a main rotation server. We're pulling it and checking with the sponsor. Apologies for the inconvenience. 06:41:09 LawrenceG has quit 06:41:09 SWP_Away has quit 06:43:45 SWP_Away has joined #emc-devel 06:43:45 LawrenceG has joined #emc-devel 06:47:11 jmkasunich is now known as jmk_sleep 06:49:53 SWP_Away has quit 06:56:48 [Global Notice] Hi all. We just experienced loss of bandwidth on a main rotation server. We've rearranged things accordingly. Apologies for the inconvenience, and thank you for using freenode. 08:22:06 SWPadnos is now known as SWP_Away 14:21:55 rayh has joined #emc-devel 14:30:35 rayh has quit 17:18:27 jmk_sleep is now known as jmkasunich 17:19:10 SWP_Away is now known as SWPadnos 17:32:23 rayh has joined #emc-devel 17:42:09 tomp has joined #emc-devel 17:42:26 good morning all 17:42:48 Hi Tom 17:46:51 hit 17:46:53 hit 17:46:56 argh 17:46:57 hi 17:48:26 hal_skeleton wont exit nicely, any ideas? 17:48:48 what does it do? (or not do) 17:49:09 tomp is running it now to report... 17:49:16 ok 17:49:54 what's a hal_skeleton? 17:50:04 sample code for writing a new driver 17:50:10 a "skeleton" driver 17:50:37 oh. something I never want to try to do again! 17:50:37 based on hal_parport, with all the parport specific code ripped out and lots of comments to aid the writer of a new driver 17:50:52 cool 17:51:13 it'sd a model, thats not a bad thing to have 17:51:20 I've tried starting a hal from realtime 17:51:25 though it is missing some (obvious) comments near rtapi_app_exit, telling you to free everything before exiting 17:51:30 and adding a few pins and sigs 17:51:30 unforch it also copied somethign I did in parport only, I attempted to make the driver work in RT an in user space 17:51:32 the 1st err is usually the kicker, it is: 17:51:40 ERROR: Module SKELETON does not exist in /proc/modules 17:51:49 so is the name? 17:51:58 is it the name? 17:52:12 tomp: it may be buggy, it hasn't been maintained 17:52:29 do a /sbin/lsmod, and see what the module name is 17:52:34 it's called SKELETON 17:52:40 then do bin/halcmd show comp, and see what its hal name is 17:52:44 comp_id = hal_init("SKELETON"); 17:52:52 thats the problem 17:52:56 ok, i just reboot, i used it to get byte sized ops 'u8' instead og _b 17:53:08 ok , I guess ed that, but the fix is??? 17:53:10 tomp, no need to reboot 17:53:45 u8 instead of bit 17:54:05 you can manually rmmod it, but you need to do that using the MODULE name, as shown my lsmod 17:54:18 not the HAL name as shown by "halcmd show comp" 17:54:58 halcmd unloadrt assumes that the HAL name is the same as the module name, and does "rmmod " 17:55:06 right - the file name is hal_skeleton.{o,ko} 17:55:19 all the other modules do indeed use the same name for hal name and module name, that one got missed 17:55:25 (and I'll fix it right now) 17:57:53 ok, su, ( sudo ng for this), then rmmod hal_skeleton (from lsmod) (works) , then rmmod others (ie: hal_lib?) 17:58:16 the others can be removed by "halcmd unloadrt all" "scripts/realtime stop" 18:00:32 ok, bugfix committed 18:01:59 thank you thank you ,that works, up&down nice (much faster to work with it now :-) 18:04:47 neat, running now,got a widget with 4 sliders under tkemc's scripts menu. it uses tcltk & hal to chg the u8 on parports, live 18:04:53 watching the leds now 18:18:13 Great. 18:18:47 This would be a really nice tuning tool inside halconfig.tcl. 18:31:43 sorry, dozing (workingon furnace sheet metal), ray, how to share it? 18:32:57 I guess I was thinking of the layout. Using sliders to set values to tuning params. 18:33:28 rather than the whole u8 to parport pins stuff. 18:33:53 ok, I'll email the .tcl, this vrsn has a 'confirm' and 'reread' btns 18:33:56 VCP should make that part easy - it's the data converter that's important :) 18:34:20 VCP? 18:34:33 the confirm btn eliminates the constant calls to halcmd, the reread is an 'oops' 18:34:35 btw, VCP is on hold right now, I'm aiming for emc-2.2 18:34:39 jmk - isn't that what it'll be called? 18:34:41 ok 18:34:46 Virtual Control Panel 18:34:48 rayh: VCP = virtual control panel 18:35:02 screen widgets that manipulate/display HAL pins 18:35:06 ah okay. 18:35:28 Is my work on halconfig a total overlap of that? 18:35:48 "buttons" work, I was working on an LED widget when I stopped and tried to focus on release stuff 18:35:58 halconfig doesn't really overlap 18:36:11 halconfig is for tuning 18:36:11 VCP is a HAL component, you can connect those widgets to others 18:36:13 rayh is now known as rayh-lunch 18:36:15 brb 18:36:22 VCP is for machine use 18:36:25 halconfig is a HAL manager, not a component 18:37:08 will you check VCP into emc2 CVS after the first release? 18:37:17 yeah 18:37:25 ok - I'd like to work on it some 18:37:33 (or at least kibbitz) 18:37:34 I suppose I could branch and check it in now 18:38:20 or just mail you the source 18:38:53 I can wait - I have a long "short list" 18:38:58 heh, ok 18:39:06 and a short "long list" ? 18:39:11 no 18:39:13 :) 18:39:16 ok then 18:39:41 can inputs get 'overloaded'? like 8 hal bits also referred to as u8? ( union like?) if in the struct? 18:39:49 no 18:39:58 tomp: right now it's pretty strict 18:40:12 you can only connect things of the exact same type 18:40:14 connecting one u8 to 8 bits would require a splitter component 18:40:28 (which is something that we have discussed) 18:40:49 ok, that why i was messin in hal_skel, to keep 'em separate 18:40:56 yep - maybe I'll do that sometime soon (unless I don't :) ) 18:41:11 we always seem to get bogged down in the details of which conversions make sense 18:41:41 I think 32->2x16, 32->4x8, 16->2x8, and 8->8x1 make sense 18:42:07 but then you have signed vs. unsigned (though unsigned only may be fine) 18:42:23 unsigned only... 18:42:37 ok, then you have the same number of "reverse" options 18:42:39 if you are thinking of the data as a value, why are you splitting it? 18:42:55 dunno :) 18:43:02 just being generic about it 18:43:52 uh, the idea would be... this byte is usually like this, and the single bit is the 'but' logic :-) (but logic ) 18:43:53 think of the reverse though - grab 4 bytes and make a 32-bit number out of them 18:44:48 when would you want to do that? when interfacing to some HW that talks in 8 bit chunks? then write a proper driver for it 18:44:56 (playing devils advocate) 18:44:59 heh 18:45:21 uh - so you can test hardware before you've written the driver? 18:46:06 yep, pre-deciding function is ok, i just didnt see any drivers using digital io in bytes 18:46:10 maybe... assuming you already have some other driver that will read the bytes, and that it reads them in the proper order (timewise) 18:46:32 a toolchanger would want to use several bits as a byte (like the one on the Mazak, I think) 18:46:43 that one I agree with 18:47:00 bits to bytes, bytes to bits 18:47:02 the CL 'script' would assure the timing 18:47:30 you then need the others so that you can have a conversion from something other than a byte to/from digital 18:47:40 yes 18:47:43 jmkasunich: how do you feel about a signal doing this? 18:47:46 for instance, a 16-bit resolver connected to zillions of bits 18:47:54 I was mostly referring to the signed ones 18:48:05 oh - losing those is no problem for me 18:48:16 SWP: what about the 16 bit resolver? 18:48:16 jmkasunich: instead of a component 18:48:22 alex: rather not 18:48:27 with parallel output 18:48:50 jmkasunich: or maybe halcmd helper functions.. 18:48:51 swp: so the 16 bits of resolver signal go to 16 individual digital ins and get combined in hal> 18:48:54 connect to 16 DIN pins, and convert to a U16 (or possibly an S16, for that matter) 18:49:03 exactly 18:49:07 because doing this will get nasty hal files (big ones) 18:49:20 disaster waiting to happen - what if the count value changes when half the inputs have been read 18:49:34 you probably need atomic operations 18:49:35 that could be a problem 18:49:42 just like the PPMC :) 18:50:48 connected to an M5I20, there wouldn't be a problem, since the read can be 32 I/Os at a time 18:51:29 rayh-lunch is now known as rayh 18:53:11 ray i opened a can-o-worms, i asked about using hal-bits as hal-bytes, suggested unions, ended up atomic & groups... 18:53:52 yeah - it's all your fault :) 18:53:52 Right variable typing has been an issue for some time. 18:54:05 I don't object to the hard typing. 18:54:24 It seems like a hal mod could be prepared that would convert 18:54:29 either way. 18:54:37 read back a little ;) 18:54:39 eight pins in to a u8 18:54:49 or a u8 to eight pins out. 18:55:02 I perfer hard typing with conversion components where needed: u8->8xbit, 8xbit->u8, u16->2xu8, 2xu8->u16, u32->4xu8, 4xu8->u32 18:55:15 prefer even 18:55:16 That works for me. 18:55:24 or, really I asked for u8 & 8 pins at same time, (cast or union) 18:55:26 that's what I said, you devil 18:55:54 (plus the 32->2x16 and 16->2x8 and 32->4x8) 18:57:02 i went with separate drivers, cuz i didnt see any such capability for >digital< io ( not encoder, float , just digi ) 18:57:03 though you're right - if you do the "direct to bits" version for the longer data types, you can get away without the other type conversions 18:57:12 SWP: I know, I was re-iterating ;-) 18:57:16 heh 18:57:30 I'm from the department of redundancy department 18:57:43 where? where? 18:57:59 is there an echo in here echoing? 18:58:09 18:58:12 heh 18:59:37 maybe a "convert" component? 19:00:12 convert would be the file, like blocks, the individial components would be named by the type of conversion they do 19:00:21 right 19:01:10 but what to call the config options? 19:01:19 U8toBits 19:01:27 BitstoU8 ... 19:01:36 could get long 19:03:17 for unsigned it 19:04:00 u8_2_bit 19:04:06 u32_2_u8 19:04:08 etc 19:04:14 for unsigned it's 3 more pairs, for signed, 3more pairs, looks like 12 types 19:04:29 (mixed case isn't used elsewhere in HAL) 19:05:13 ok 19:05:39 and use the same name in the config string as the component name in HAL? 19:06:12 you mean the insmod arg? yes 19:06:17 yep - ok 19:06:19 u8_2_bit=3 19:06:25 if you want 3 of them 19:06:34 just like blocks.c 19:06:52 do we need more scenarios before coding? like see what is practically needed? 19:07:24 we're disacussing how to spell the names already 19:07:29 in that case u8_2_bit would be the component name? 19:07:50 if so I like ddt better. 19:08:05 rayh ? 19:08:06 maybe use "to" instead of "2" 19:08:21 what was ddt? old memories stirred 19:08:31 I'm just thinking of how the pins look with show. 19:08:31 derivative (d/dt) 19:08:32 ddt is the differentiator block in blocks.c 19:08:38 and others 19:09:08 but what did you mean ray? 19:09:16 u8_to_bit, 8utobit? 19:09:26 oops 19:09:28 a ddt pin might be ddt.0.in 19:09:31 u8_to_bit, u8tobit? 19:09:46 u8->bit ;) 19:09:47 actually, just calling them "byte.n" would make some sense - you would have pins like "byte.0.bit-0-out" 19:09:54 pins would be u8_to_bit.0.bit.0.in 19:10:14 which way does "byte" convert? 19:10:16 err - u8_to_bit.0.bit.0.out 19:10:28 oops, right 19:10:30 or call them u8 in that case 19:10:31 hierarchical? byte.0 and byte.0.bit.o? 19:10:35 u8.0.bit.0.out 19:10:41 yeh yeh 19:10:48 bits.0.u8.out 19:10:56 bits.1.u16.out 19:10:58 ... 19:11:06 * rayh gets lost in all the possibilities. 19:11:18 (possibly make bits8 and bits16, for cfg purposes) 19:11:21 now you know why we haven't made that component yet 19:11:24 heh 19:11:26 now no need for casts unions overloading, it's poppa and baby-brother 19:11:36 and we haven't even gotten into float / int conversions yet 19:13:22 float/int conversions make little sense to me 19:13:39 floats are for inherently analog (continuous) signals 19:13:44 ints are quantized 19:13:54 yep 19:14:05 but then there's the R-2R ladder you might connect to the parport 19:14:23 we do have drivers that convert from int to float, like the encoders that convert raw counts to position 19:14:29 and it would be a heck of a lot easier to not write a special driver for it 19:14:38 knob detent at posn 5 goes to spindle speed which is a float, spindle speed float goes to led speed5 19:14:46 they also have scale and offset, because analog -> quantized conversions probably involve scaling 19:14:51 yes 19:15:02 you can't just directly convert 19:15:20 ther eare scale/offset float compe, but no way to get that to int (right?) 19:18:12 ok, you win ;-) I'm convinced that there is use for a float to/from s32 conversions 19:18:34 heh 19:18:43 u32 to float would also work, float to u32 has a problem if the float is negative 19:19:01 or >MAXINT 19:19:20 there would probably be max and min params anyway 19:48:37 Hey I need a quick way to change a scientific notated value to decimal? 19:48:47 bc says it won't 19:49:10 see lots of stuff that will convert decimal to scientific notation. 19:49:26 in tcl? 19:49:36 Yes. 19:49:57 it should be able to interpret the scientific notation as a number, then use format to print it 19:50:07 right now it's a [split $string +] 19:50:12 or does it not "get it" with scientific notatipn? 19:50:23 notation 19:50:26 every variable is a string. 19:50:48 eval? 19:50:55 variables get converted to numbers to do math, right? 19:51:10 [ eval $variable + 0 ] 19:51:19 they must, or var = var + 1 would give a long string of '1's 19:52:29 not eval, expr 19:52:37 right. 19:53:08 within expr it can treat variables as having value 19:53:18 set foo 1.4e4 puts [ expr $foo ] gives 14000.0 19:56:48 yep set bar [expr $foo] 20:14:23 Ah I'll try that. Thanks guys. 20:44:50 has anyone heard of PCX/DSP from MEI(Motion Engineering Inc, US) 20:50:19 jmkasunich: you might be interested by this.. 20:50:33 it's an 8-axis control board with a fast DSP on it 20:50:59 they provide libraries for it (inthe form of .dll's, I thikn) 20:51:03 think 20:51:19 right, and I have an user asking now if it can be modified for emc2 20:51:30 to access directly the dac's adc's etc 20:52:28 oh.... 20:52:31 * jmkasunich hides 20:52:37 they havea Linux port - have you found out if it's open source? 20:53:00 hang on, I have some sources he sent me 20:54:40 it looks like you have to pay for the libraries 20:55:27 sounds like a pmac, maybe ressurrect the old emc wrapper for pmac 20:55:34 I dunno why you would want to pay for a fast DSP only to bypass it and access the I/O directly 20:55:54 it seems like it was designed to handle the RT load so a non-RT OS could be used to do the rest 20:55:54 and it seems to do S-curve profile 20:55:59 you can make a HAL pid block that just sends the params on to the DSP 20:56:11 it does have PID in it ;) 20:56:24 though a 40MHz DSP may not give any speed advantage 20:56:54 To create a motion sequence, the DSP executes a series of ?frames? that are generated by MEI C library functions and sent from the host. Each frame is an array of 20 words that contain position, velocity, acceleration, jerk, I/O status, and trigger information. 20:57:13 sounds fairly ideal, actually ;) 20:57:26 I bet it would - a 40MHz DSP running an optimised fixed point algorithm will blow the socks of a GHz class general purpose CPU when it comes to latency and jitter 20:57:36 true 20:57:45 a 1MHz DSP might, come to think of it 20:58:13 at least in the jitter category 20:59:16 well, I'm not sure what to advise this user.. 20:59:33 depends on their skill level and motivation 20:59:47 " I thought I will just ask you before I start to involve someone 20:59:47 to write a driver, if first of all, is it possible to use this board 20:59:47 and just access the low lovel encoder, dac and i/o. 20:59:47 And second, may it be worth it, or is it better or cheaper to buy a m5i20 board for around 400 us dollars? 20:59:49 maybe that it wold be possible to write a driver if the specs to the board or the library are made available to the developers ... 20:59:56 if they have coding skills and the dev tools for the DPS, they can certainly make it work 21:00:22 it's a question of having a proprietary driver for emc, I think 21:00:40 I might be willing to write the HAL driver (PC side), but I suspect it will need changes to the DSP code as well, I won't do that 21:00:41 since it looks like the programming specs aren't generally available 21:02:26 hmm, there are some functions available to access low level stugg 21:02:29 stuff 21:02:45 int FNTYPE dsp_port_address(int port, P_INT address) ; 21:02:45 int FNTYPE pcdsp_get_io(PDSP dsp, int port); 21:02:45 int FNTYPE pcdsp_set_io(PDSP dsp, int port, int value); 21:02:45 int FNTYPE pcdsp_init_io(PDSP dsp, int port, int config); 21:02:55 for me at least, any work on a driver would have to wait until after the emc-2.0.0 release 21:02:57 and so on... 21:03:06 but it certainly wouldn't be GPL 21:03:20 then I won't work on it 21:03:38 right 21:03:39 (unless they wanted to pay me lots of money ;-) 21:03:52 and me too .. for hooking them up :D 21:04:08 you do need a manager afterall :D 21:04:20 ;-) 21:05:29 * jmkasunich tries to figure out how to detect and report RT overruns 21:07:11 I just discovered something interesting... the BDI-2.18 box on the compile farm does NOT have sudo installed or configed 21:07:23 so the new emc2 run script that uses sudo... won't work right 21:10:23 oh.. 21:10:37 but it would I think.. 21:10:45 sudo gets found by configure 21:10:57 and it replaces teh $SUDO with what it finds 21:11:09 so if no duso is found $SUDO should be empty 21:11:22 s/duso/sudo/ 21:14:09 why do I get this -- motmod u8 R- 0 (00) traj.active_tc 21:14:10 m 21:14:12 so the user just needs to do 'su -c "scripts/emc myconfig"' 21:14:41 all the u8 replies give two values 21:14:50 in that column. 21:14:51 using -s? 21:14:55 decimal and hex 21:14:56 Yes 21:15:02 0 is the simple (hex) is other 21:15:02 maybe I missed that 21:15:06 ah. 21:15:25 -s should fix that 21:15:31 give you decimal only 21:15:49 That would be more consistent with the rest and make my parsing easier. 21:16:11 what command gave you that line? show param? 21:16:49 yes 21:17:02 bin/halcmd -s show param 21:18:42 hmm - are you absolutely positive that -s was in there? 21:18:51 doing the same thing here 21:18:55 all the u8s 21:19:05 stepgen s32 -W 0 stepgen.2.counts 21:19:14 works 21:19:20 only u8 is odd. 21:19:45 originally only u8 displayed the hex (the others it would have been too long) 21:20:07 but SWP added the -s specifically to eliminate the hex (among other script friendly things) 21:20:15 and I thought it was tested and working weeks ago 21:20:15 print_param_info has a separate case for the script mode 21:20:51 data_value must not be printing decimal 21:21:17 that's teh thing - when you get the name, there's no way to get two numbers from that printf 21:21:36