Wednesday, 2014-02-05

*** tpb has joined #timvideos00:00
mithropuck: I've updated the documentation a bit03:02
mithroCarlFK: regarding the " "Oh man, are people actually using that old code?  This is no longer maintained and has been superseded by the fxload app in the libusbx directory." -- Turns out he removed functionality needed in the latest versions03:02
tpbTitle: pbatard/fxload ยท GitHub (at
puckmithro:  ta, I need to head off now, tomorrow is a holiday, I'll be back at it on Friday03:22
mithrothaytan: just fyi04:02
tpbTitle: Pastebin: xFdDo8E3 | IRCCloud (at
thaytanmithro, seems a bit off, eh?04:07
mithroWell, it's of the order of 50-40 microseconds04:08
mithro(at worse)04:09
mithroI'm trying to remember my signal processing to see if there is something that could give me a better understanding of the jitter04:12
thaytanI really suspect it'll just be timing error taking the measurement04:16
thaytanalthough it's interesting that the average is about 10 uS shy04:17
thaytanI wonder why / how it reports that 16661 value04:17
mithrothaytan: a theory is that is how long it takes to service the interrupt04:18
thaytanright, but somehow it still reports a different value for the internal framerate estimate04:18
mithrowhat do you mean?04:22
thaytanmithro, that it reports 16661uS for the internal measure of the framerate04:29
thaytanbut the average from measuring it yourself is 1665104:29
thaytan10 uS shorter04:30
thaytanso how is libGL measuring it?04:30
mithrothaytan: I assume it is just passing up what the monitor is saying it does04:30
thaytanit can't get it direct from the monitor - it's measured against the system clock on the PC04:30
thaytanwould have to drill into libGL to find out04:31
thaytanor ask keithp or anholt04:31
mithroyeah, I just joined #intel-gfx to see if someone there could help04:34
mithrothaytan: you assume it is doing the right thing :P04:34
*** mproctor has quit IRC04:39
mithrothaytan: it probably just reads the edid information and reports that :P04:40
thaytanmithro, it is at least internally consistent - 4700000/78309 == 60.018644Hz == 1661.4893 uS per frame04:40
thaytanmithro, that doesn't make sense to me04:41
thaytanif the monitor knew it wasn't doing exactly 60fps... it'd be able to adjust slightly so it was04:41
mithromy code could be wrong :P04:42
thaytanthat I believe ;)04:42
tpbTitle: Pastebin: Pp0x1M56 | IRCCloud (at
thaytandamnit, don't make me go read Mesa code :-P04:43
thaytanit is reporting based on the system dot clock taken from XF86VidModeGetModeLine04:47
mithrothaytan: I'm chatting with keithp now05:02
thaytanoh good05:03
thaytanit makes sense to me a bit more now05:03
thaytanafter learning GetMscRateOML comes from the dotclock05:04
mithrowhat is "the dotclock" ?05:06
thaytanit's the crystal on the graphics chip that clocks out the pixels05:06
thaytanit gets set to the closest supported value to output the requested video mode05:07
thaytanwhich can't always hit the ideal value for a given framerate05:07
CarlFKmithro: pretty sure the functionality is there, just under a different option.  like -d vs -D (we have had this discussion before.. maybe I should open an issue?07:50
mithroCarlFK: I'm pretty sure it was missing07:53
CarlFKwhats the functionality?07:53
mithroThe udev script needs the ability to specify a USB device in /sys for when you have two devices with the same vid:pid (IE you plug in two devices)07:54
mithroIt may have been added since I last looked.07:56
CarlFK$ djtgcfg prog -d Atlys -i 0 -f hdmi2usb.bit07:56
CarlFKpretty sure it is now -D07:56
mithroFx load07:56
CarlFKbut that may be a hex07:56
CarlFKoh.. back to docs..07:56
CarlFKsee ya07:56
*** CarlFK1 has joined #timvideos12:06
*** CarlFK has quit IRC12:06
*** CarlFK1 is now known as CarlFK12:07
*** CarlFK has quit IRC12:16
*** CarlFK has joined #timvideos12:18
*** mproctor has joined #timvideos21:22
mithroCarlFK: heyo23:46
CarlFKmithro: perfect timing23:46
CarlFKI am on a box with 24 cores (not sure if they are real, or 12*2.. whatever, don't care, call it 24.23:47
CarlFKI need to kick of 22 instances of my encoder (webm isn't multi tread)23:48
CarlFKI think I did something like that with bash script and screen -d -m .. but couldn't figure it out a few days ago23:48
CarlFKany suggestions ?23:49
CarlFKthat did come to mind23:49
mithrosubprocess and queue should work well23:50
mithroCarlFK: I could write you up something which just takes the command lines on stdin and reports when they finish on stdout23:51
CarlFKI think I want them running in a screen or tmux23:52
CarlFKI don't trust things will run well enough, and if it fails I want to be able to attach to the session to see what is going on23:53
CarlFKlike I just bumped into a unicode problem that somehow eluded your fix23:54

Generated by 2.12.1 by Marius Gedminas - find it at!