2017-06-10T00:04:37Z rpg joined #sbcl
2017-06-10T00:07:48Z |3b|: looks like builds if i add :margin-size #x8000 to !gencgc-space-setup
2017-06-10T00:08:37Z |3b| has no idea if that is the right thing to do though
2017-06-10T00:13:40Z jack_rabbit joined #sbcl
2017-06-10T00:24:16Z cromachina joined #sbcl
2017-06-10T00:27:03Z |3b| can't tell if it had any effects on the tests or not
2017-06-10T00:32:32Z rumbler31 joined #sbcl
2017-06-10T00:36:43Z rumbler31 quit (Ping timeout: 240 seconds)
2017-06-10T00:52:01Z jack_rabbit quit (Ping timeout: 246 seconds)
2017-06-10T00:52:50Z jack_rabbit joined #sbcl
2017-06-10T01:11:48Z malm joined #sbcl
2017-06-10T01:33:17Z rumbler31 joined #sbcl
2017-06-10T01:37:38Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-10T02:34:06Z rumbler31 joined #sbcl
2017-06-10T02:38:32Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-10T03:34:48Z rumbler31 joined #sbcl
2017-06-10T03:39:05Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-10T03:39:48Z Bike quit (Ping timeout: 260 seconds)
2017-06-10T04:35:37Z rumbler31 joined #sbcl
2017-06-10T04:40:02Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-10T05:09:59Z rpg quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
2017-06-10T05:15:21Z rumbler31 joined #sbcl
2017-06-10T05:19:38Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-10T05:46:20Z gingerale joined #sbcl
2017-06-10T06:16:11Z rumbler31 joined #sbcl
2017-06-10T06:20:23Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-10T06:44:31Z shka_ joined #sbcl
2017-06-10T07:04:14Z ym quit (Ping timeout: 268 seconds)
2017-06-10T07:36:17Z shka_ quit (Ping timeout: 260 seconds)
2017-06-10T07:36:52Z shka_ joined #sbcl
2017-06-10T07:40:04Z oleo joined #sbcl
2017-06-10T08:06:03Z shka_ quit (Ping timeout: 260 seconds)
2017-06-10T10:06:11Z rumbler31 joined #sbcl
2017-06-10T10:10:30Z rumbler31 quit (Ping timeout: 240 seconds)
2017-06-10T10:33:43Z rumbler31 joined #sbcl
2017-06-10T10:37:57Z rumbler31 quit (Ping timeout: 245 seconds)
2017-06-10T10:48:23Z salv0 joined #sbcl
2017-06-10T10:50:17Z oleo quit (Quit: irc client terminated!)
2017-06-10T10:51:11Z Bike joined #sbcl
2017-06-10T10:51:25Z oleo joined #sbcl
2017-06-10T11:17:34Z shka_ joined #sbcl
2017-06-10T11:50:21Z rumbler31 joined #sbcl
2017-06-10T11:54:44Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-10T12:01:06Z rpg joined #sbcl
2017-06-10T12:05:40Z rpg quit (Ping timeout: 255 seconds)
2017-06-10T12:29:21Z Merv_ quit (Remote host closed the connection)
2017-06-10T13:06:44Z rumbler31 joined #sbcl
2017-06-10T13:11:13Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-10T13:24:47Z ^eddy joined #sbcl
2017-06-10T13:25:03Z AndreasO joined #sbcl
2017-06-10T13:35:21Z ^eddy: If one wished to make a SBCL thread immune to GC, would it make sense to have a declaim directive that omits the safepoint from functions called by said thread?  And, hopefully, forbids any allocations that are not dynamic-extent?  A naive implementation obviously would imply safety 0. ;-)
2017-06-10T13:36:59Z ^eddy: I STFWed, and am okay with the "just use C" approach.  However, it appears that I'm not the only one who would rather stick with SBCL in near-RT work; perhaps it makes sense to try creating a dirty hack for others who like to live on the wild side?
2017-06-10T13:47:33Z rumbler31 joined #sbcl
2017-06-10T13:51:49Z rumbler31 quit (Ping timeout: 246 seconds)
2017-06-10T14:07:51Z cromachina_ joined #sbcl
2017-06-10T14:09:05Z cromachina quit (Ping timeout: 240 seconds)
2017-06-10T14:10:24Z AndreasO quit (Quit: Found more important stuff than irc!)
2017-06-10T14:13:38Z karswell` joined #sbcl
2017-06-10T14:36:08Z flip214: ^eddy: anything you can come up with would be interesting.
2017-06-10T14:36:09Z minion: flip214, memo from easye: First batch of edits to the ever finer Manual from your review <http://abcl.org/trac/changeset/15067>.
2017-06-10T14:36:38Z karswell` quit (Remote host closed the connection)
2017-06-10T14:37:22Z karswell` joined #sbcl
2017-06-10T14:40:46Z ^eddy: flip214: Thanks. I'm fairly new to SBCL guts (although very comfortable in asm), and want to avoid The Wrong Way (TM).
2017-06-10T14:41:39Z flip214: ^eddy: I'm using sbcl on an RPi2 for PWM (LED lights, each timeslot ~40msec), and don't have any issues.
2017-06-10T14:41:57Z flip214: guess that the few allocations I'm doing can be reaped in no time.
2017-06-10T14:42:32Z flip214: or am I using 10msec now?  can't remember.
2017-06-10T14:42:40Z ^eddy: We've been using LabVIEW *shudder* at work for some robotics-esque work.  There are a few control loops that presently run down to ~15 ms, but faster is desirable.
2017-06-10T14:43:05Z flip214: well, on ARM there's no thread support for SBCL...
2017-06-10T14:43:28Z ^eddy: Management is not interested in true realtime or embedded; I don't have _that_ much latitude.  But I do have enough to choose the software platform.
2017-06-10T14:43:39Z flip214: so I've only got a process with one thread that does PWM via usleep(), and every 10msec checks for new data  (via a file in shmfs)
2017-06-10T14:44:21Z flip214: well, if you're (and your platform!) are okay with busy-waiting, down to some 20-100 usec should be possible.
2017-06-10T14:45:06Z ^eddy <-- stuck with Windows on x86-64.  Can use longer sleep() when it makes sense, then sleep(0) to "spin" nicely.
2017-06-10T14:45:34Z flip214: ouch.
2017-06-10T14:45:51Z ^eddy: Re ARM: I'd love to get Cortex R in house.
2017-06-10T14:45:56Z flip214: which windows?
2017-06-10T14:45:59Z ^eddy: Win7
2017-06-10T14:46:10Z ^eddy: Plain old desktops.
2017-06-10T14:46:39Z flip214: for robotics? a "big tower" with feet?
2017-06-10T14:47:08Z flip214: remove the virus scanner ;)
2017-06-10T14:47:28Z ^eddy: I'm generalizing/simplifying when I say robotics.  It's for configurable testing of manufactured mechanical parts.  Read sensors, decide what to do, output control signals.
2017-06-10T14:47:48Z flip214: oh, okay. still needs to react within a few msec, right.
2017-06-10T14:48:11Z flip214: and windows is fixed? no way around that?
2017-06-10T14:48:19Z rumbler31 joined #sbcl
2017-06-10T14:48:27Z ^eddy: for now, at least
2017-06-10T14:48:42Z ^eddy: It's taken me nearly three years to get them moving away from LabVIEW.
2017-06-10T14:48:45Z flip214: because linux with a preempt kernel would get you down to ~6usec, according to an article in iX (heise.de)
2017-06-10T14:49:07Z ^eddy: And the quantum is configurable on *ix.  Yes.
2017-06-10T14:49:16Z flip214: well, just put linux with LXQT or so on the box, and claim that to be "the realtime edition of windows"
2017-06-10T14:49:27Z flip214: ;)
2017-06-10T14:49:39Z Bicyclidine joined #sbcl
2017-06-10T14:49:39Z ^eddy: In a previous lifetime, I ran FreeBSD-based routers with I forget what quantum to minimize jitter.
2017-06-10T14:49:44Z ^eddy: heh
2017-06-10T14:49:52Z ^eddy: "It _looks_ like a duck, right?"
2017-06-10T14:50:49Z flip214: I'd use linux -- for example, if some unexpected input comes in (image from camera can't be interpreted etc.)
2017-06-10T14:51:08Z flip214: it's just a small RUN-PROGRAM line to send that picture via email to the developer...
2017-06-10T14:51:20Z flip214: yeah, well, in windows you can do that too... install cygwin or mingw
2017-06-10T14:51:39Z flip214: so that won't be a good argument
2017-06-10T14:51:41Z ^eddy: Still the same underlying Windows OS, though.
2017-06-10T14:51:45Z ^eddy: yeah
2017-06-10T14:52:22Z ^eddy: So... I began digging into SB internals, lest I have an embarassing round of Java-style stutter when the GC hits.
2017-06-10T14:52:28Z Bike quit (Ping timeout: 240 seconds)
2017-06-10T14:52:56Z rumbler31 quit (Ping timeout: 255 seconds)
2017-06-10T14:53:17Z flip214: ^eddy: well, just write the program to do as few consing as possible...
2017-06-10T14:53:39Z flip214: or put explicit GC calls when appropriate (just after activating some output signal?)
2017-06-10T14:53:44Z flip214: https://stackoverflow.com/questions/628057/how-to-set-processor-affinity-on-an-executable-in-windows-xp
2017-06-10T14:53:50Z ^eddy: *node*
2017-06-10T14:54:00Z ^eddy: err, *nod*
2017-06-10T14:54:11Z flip214: pin the SBCL process to one cpu, and all other process to the other CPUs....
2017-06-10T14:54:21Z flip214: that might help a bit
2017-06-10T14:54:36Z flip214: only the locking stuff in syscalls will still be unpredictable
2017-06-10T14:54:48Z ^eddy: For a first pass, at least.
2017-06-10T14:54:49Z flip214: what kind of IO are you using?
2017-06-10T14:55:00Z ^eddy: async and IOCP
2017-06-10T14:55:09Z ^eddy: of course :)
2017-06-10T14:55:14Z flip214: no, sorry.
2017-06-10T14:55:20Z flip214: PCIe card with GPIO?
2017-06-10T14:55:26Z flip214: USB2serial and control lines?
2017-06-10T14:55:29Z flip214: parallel port?
2017-06-10T14:55:32Z flip214: USB camera?
2017-06-10T14:55:40Z ^eddy hangs his head and admits "USB" (but not camera)
2017-06-10T14:55:57Z flip214: how many output pins do you need?
2017-06-10T14:56:11Z flip214: (and what kind of input)?
2017-06-10T14:56:33Z ^eddy: It varies.  We generally have 1-4 analog out, up to 12 analog in, and a few digital.
2017-06-10T14:57:21Z flip214: hmmm.... what kind of output frequency, and input sample speed?
2017-06-10T14:57:40Z flip214: what input precision? 8bit, 10, 12, 16, 24 bits?
2017-06-10T14:58:02Z ^eddy: My predecessor tried to use a Cypress PSoC 1, which left them with a sour taste for embedded systems.
2017-06-10T14:58:41Z flip214: well, depending on the exact needs, you might be better off with something that can drive the IO "more directly"...
2017-06-10T14:58:42Z ^eddy: generally IO at <= 111 Hz
2017-06-10T14:59:09Z flip214: an RPi with MCP3008 (8 -Channel 10-Bit ADC) via SPI, for example
2017-06-10T14:59:12Z ^eddy: when faster sampling is needed, the hardware takes care of that, and a bulk read at slower interval is fine
2017-06-10T14:59:39Z flip214: ah, so that variable latency from USB might not actually matter that much
2017-06-10T15:00:45Z ^eddy: The high-speed sampling is to detect things like switch bounce.  I've already suggested having an analog comparator drive a latch, rather than just hoping that we catch the bounce "in the act".
2017-06-10T15:01:13Z flip214: uh, I'd just debounce in hardware!?
2017-06-10T15:01:26Z ^eddy: In short: There is much that I would like to change. I'm taking progress where I can get it. I want to have SBCL make a good showing in near-RT. :)
2017-06-10T15:01:32Z flip214: and "111 Hz" is a strange requirement for debounce ;)
2017-06-10T15:01:53Z flip214: yeah, right!    sorry, gotta go. please ping me for any news or questions.... sounds interesting!
2017-06-10T15:02:05Z ^eddy: 111 Hz = 9 msec = just above some hardware that needs 8 msec to "do its thing" before being read
2017-06-10T15:02:29Z ^eddy: Cool.  I'll keep digging around SBCL guts, and see where I end up.
2017-06-10T15:02:37Z ^eddy: Catch you later.
2017-06-10T15:02:46Z flip214: well, _that_ reasoning makes me go "urgh", sorry ;)
2017-06-10T15:02:53Z flip214: yeah, keep me posted, please!
2017-06-10T15:03:09Z ^eddy nods
2017-06-10T15:35:28Z ^eddy discovers inhibit-safepoints in compiler/policies.lisp
2017-06-10T15:38:31Z nyef` quit (Ping timeout: 255 seconds)
2017-06-10T15:46:17Z rumbler31 joined #sbcl
2017-06-10T15:50:27Z rumbler31 quit (Ping timeout: 245 seconds)
2017-06-10T17:06:27Z Merv_ joined #sbcl
2017-06-10T17:14:21Z AndreasO joined #sbcl
2017-06-10T17:21:40Z AndreasO quit (Quit: Found more important stuff than irc!)
2017-06-10T17:38:30Z ^eddy quit (Ping timeout: 240 seconds)
2017-06-10T17:40:50Z Merv_ quit (Ping timeout: 240 seconds)
2017-06-10T17:52:26Z flip214: christoph_debian: would you please package sbcl for arm64, too?
2017-06-10T17:53:04Z flip214: thanks a lot!
2017-06-10T18:05:04Z gingerale quit (Remote host closed the connection)
2017-06-10T18:25:47Z christoph_debian: flip214: exists on arm64 for quite a while already
2017-06-10T18:27:03Z christoph_debian: since january 2016 it seems
2017-06-10T18:29:04Z flip214: christoph_debian: ah yes, sorry.... seems I can't read.   https://packages.debian.org/search?arch=arm64&keywords=sbcl
2017-06-10T18:29:07Z flip214: shows it.
2017-06-10T18:29:19Z flip214: or my eyes auto-corrected "arm64" to "amd64"
2017-06-10T18:37:02Z Merv_ joined #sbcl
2017-06-10T18:39:36Z igajsin joined #sbcl
2017-06-10T18:41:44Z Merv_ quit (Remote host closed the connection)
2017-06-10T18:41:59Z Merv_ joined #sbcl
2017-06-10T19:02:27Z christoph_debian: (-:
2017-06-10T19:45:34Z igajsin quit (Ping timeout: 255 seconds)
2017-06-10T19:49:33Z aeth: now someone needs to come up with an "amr64" instruction set.
2017-06-10T20:00:11Z igajsin joined #sbcl
2017-06-10T20:06:57Z fiddlerwoaroof: (defmethod foo ((bar keyword))) hangs in sbcl 1.3.18 and 1.3.17.88-a40115dc3
2017-06-10T20:10:51Z fiddlerwoaroof: It works in 1.3.14 but hangs in 1.3.16 as well (testing the various versions of sbcl I have installed)
2017-06-10T20:17:14Z fiddlerwoaroof: (I know that KEYWORD isn't a class, but it shouldn't hang)
2017-06-10T20:18:20Z jack_rabbit quit (Ping timeout: 246 seconds)
2017-06-10T20:31:07Z fiddlerwoaroof: The backtrace looks like: http://paste.lisp.org/+7GW8
2017-06-10T21:28:11Z angavrilov quit (Remote host closed the connection)
2017-06-10T21:33:30Z shka_ quit (Ping timeout: 240 seconds)
2017-06-10T21:45:50Z aeth quit (Ping timeout: 246 seconds)
2017-06-10T21:47:54Z aeth joined #sbcl
2017-06-10T21:59:03Z slyrus quit (Read error: Connection reset by peer)
2017-06-10T21:59:33Z slyrus joined #sbcl
2017-06-10T22:06:56Z stassats joined #sbcl
2017-06-10T22:10:41Z fiddlerwoaroof: Using git bisect, I think the first bad commit is 810ccc869678de3df38aac41df8ca39b47388f93
2017-06-10T22:17:48Z slyrus__ joined #sbcl
2017-06-10T22:18:27Z slyrus_ quit (Ping timeout: 240 seconds)
2017-06-10T22:18:30Z slyrus quit (Ping timeout: 240 seconds)
2017-06-10T22:18:43Z slyrus__ is now known as slyrus
2017-06-10T22:18:59Z slyrus_ joined #sbcl
2017-06-10T22:21:00Z fiddlerwoaroof: nevermind, that isn't right
2017-06-10T22:26:36Z slyrus__ joined #sbcl
2017-06-10T22:26:46Z slyrus quit (Ping timeout: 268 seconds)
2017-06-10T22:29:21Z slyrus__ is now known as slyrus
2017-06-10T22:33:42Z slyrus quit (Ping timeout: 240 seconds)
2017-06-10T22:36:16Z slyrus joined #sbcl