05:59 <sb0> rjo, I have an idea for DDS write injection on the shared bus. the RTIO core could forward the time left before the next kernel-programmed event to the PHYs, and then the DDS PHY uses that to determine if it has time to complete a full write from the injection interface

05:59 <sb0> this is complicated because of 1) the guard buffer in the RTIO core 2) the problem of restoring the kernel DDS register values when the override is disabled

05:59 <sb0> but probably doable

06:02 <sb0> there's also the problem of dealing with the "no event" case...

06:03 <sb0> well, that one and the guard buffer problem can be solved by accounting for a potential DDS write in the guard time.

06:03 <sb0> and proceed with the injected writes when there is no kernel event

10:48 <sb0> http://www.bizjournals.com/sanjose/news/2015/02/11/tabula-to-shut-down-120-jobs-lost-at-fabless-chip.html

10:50 <sb0> 'another in a long line of tech startups with good technology that just didn't make it into the market because of timing or the market dynamics' - and software that doesn't work, perhaps

10:50 <sb0> ISE is a blessing compared to the tabula software

12:07 <GitHub72> [artiq] sbourdeauducq pushed 2 new commits to master: http://git.io/vIgCU

12:07 <GitHub72> artiq/master 276a178 Sebastien Bourdeauducq: gui: add TTL override label

12:07 <GitHub72> artiq/master b2af0f6 Sebastien Bourdeauducq: soc,runtime: support TTL override

12:39 <travis-ci> m-labs/artiq#199 (master - 276a178 : Sebastien Bourdeauducq): The build passed.

12:39 <travis-ci> Build details : http://travis-ci.org/m-labs/artiq/builds/66038117

13:11 <sb0> rjo, what should the DDS monitor/injection panel show? just frequency, or more?

14:18 <MiW> hiya

14:18 <MiW> someone told me Sébastien got wifi working on a baytrail tablet?

14:19 <MiW> id love to hear more if anyone knows about this

14:24 <sb0> hi MiW

14:24 <sb0> yeah

14:24 <MiW> hi sb0

14:25 <sb0> which tablet/wifi chipset?

14:25 <MiW> ive been working on an arch install on a hp omni 10

14:26 <sb0> the one i have is lenovo tablet 10, with a broadcom sdio wifi chip

14:26 <MiW> yeah its probalby the same actually

14:27 <sb0> on this tablet, for the wifi you just need to use a recent kernel (4.1-rcX) and the latest firmware that got committed to linux-firmware

14:27 <sb0> and the wifi works fine.

14:27 <sb0> there are many other bugs, though

14:27 <MiW> oh mm nothing special then

14:27 <sb0> since recently, no

14:27 <MiW> i built a few kernels for it a few months ago but no luck

14:27 <sb0> < 4.1 has packet delay/power management bugs, but those got fixed apparently

14:28 <sb0> < 4.0 have sdio bugs that prevent the wifi chipset from being detected at all

14:28 <sb0> ...and all kernels have sdio bugs that crash the eMMC, which is the most annoying bug this tablet has

14:29 <MiW> yeah, the emmc thing is a pain

14:29 <sb0> also, there are acpi bugs: no AC adapter, no battery, and suspend just turns the screen off and freezes the machine completely

14:30 <MiW> i did not have the same experience on my omni with that

14:30 <sb0> which creates a nasty combination with systemd, which automatically suspends your laptop when on battery and no AC

14:30 <sb0> *on low battery

14:30 <MiW> yikes

14:30 <sb0> the workaround is to build the kernel with the suspend support compiled out

14:31 <MiW> im going to play with it some more now i know that wifi can work on that chipset

14:32 <sb0> also, the broken AC adapter driver creates thousands of bogus events per second that cause userland issues and waste CPU cycles, so you want to compile out that one as well

14:32 <MiW> this is on 4.1 yeah?

14:32 <MiW> i never had such issues on 3.19 or 4.0

14:32 <MiW> just no wifi and crackling sound

14:32 <sb0> recent 4.0.X kernels have the bug.

14:32 <MiW> oh. :-)

14:32 <sb0> with older one, just no AC adapter detected

14:33 <sb0> the bug appeared with a patch that changes the way interrupts are handled, and that patch is required to make the wacom digitizer work

14:35 <MiW> wow this got worse quickly

14:35 <MiW> i have a semi working tablet 3 months ago

14:35 <MiW> now its ruined

14:35 <MiW> hah

14:35 <sb0> ?

14:35 <MiW> like, the builds i was playing with 3 months ago were OKish just no sound/wifi

14:35 <sb0> the AC adapter driver never worked anyway

14:36 <sb0> at least on mine

14:36 <MiW> ok

14:36 <MiW> you have inspired me to do some more work on this

14:36 <sb0> do you have the battery monitor/AC adapter detection working?

14:37 <MiW> I dont recall. it worked on battery or AC tho i remember

14:37 <sb0> the AC adapter driver only lets userland know if the computer is on AC or not

14:38 <sb0> the battery will charge etc. even without the driver

14:44 <MiW> yes thats right

15:23 <ysionneau> sb0: since the pyqtgraph didn't respond in a reasonable delay (even after your issue on github + my private emailing) could you fork his repo in the m-labs github account? (I don't have the right to do it)

15:23 <ysionneau> pyqtgraph maintainer*

15:24 <ysionneau> I propose we can go back to using his upstream repo when he fixed the trivial issue

15:48 <ysionneau> hummm, seems max output frequency of PLL (for -2 speed grade device) is 933 MHz

15:48 <ysionneau> so I don't think I can make the SerDes work @ 125 MHz x 8 (= 1GHz)

15:49 <ysionneau> I would need to lower a bit the rio frequency to lower than 125 MHz

15:50 <ysionneau> or maybe I can put the SerDes in DDR mode and use a 500 MHz clock

21:29 <rjo_> sb0: yes the rtio side of the dds-bus fifo could accept injected writes that can be squezed into the (enlarged) guard window.

21:30 <rjo_> dds should just show frequency. imho the other things (phase, phase mode, amplitude) are of at most secondary importance.

21:31 <rjo_> ysionneau: the ddr mode looks like the right thing.

21:31 <ysionneau> I was able to generate a bitstream which met the timing constraints with the ddr mode

21:32 <rjo_> *dds injmon should ..

21:34 <rjo_> ysionneau: nice. in case you don't have a fast scope: i had verified that the ventilator hires dtc actually does what it is supposed to do. clock was 8*83 MHz.

21:39 <ysionneau> good!

21:40 <ysionneau> I took some ideas from your code

21:40 <ysionneau> but I had to change a bit the SerDes instanciations since 7 series use different serdes blocks :/

21:41 <ysionneau> (and I don't have any scope nor any 7 series FPGA ^^)

21:43 <rjo_> ack. it would be nice to have the s6 version as well. pipistrello might see quite a bit use with the old hardware in the lab.

21:45 <ysionneau> one it will work on 7 series, I don't think it will be hard to port on the S6

21:45 <ysionneau> once*

22:53 <GitHub49> [artiq] fallen pushed 1 new commit to master: http://git.io/vIw94

22:53 <GitHub49> artiq/master 92999d0 Yann Sionneau: sync_struct: test append, insert, delitem and pop actions

23:21 <travis-ci> m-labs/artiq#200 (master - 92999d0 : Yann Sionneau): The build passed.

23:21 <travis-ci> Build details : http://travis-ci.org/m-labs/artiq/builds/66131154