azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | | Logs:
<azonenberg> Sooo my attempts to date at making a cleaner tree store have been... problematic
<azonenberg> but this is top priority for me as it's blocking stuff for $dayjob
<azonenberg> it seems like ListStore is a lot more efficient than TreeStore
<azonenberg> but obviously it doesnt handle child nodes at all
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 252 seconds]
Degi_ is now known as Degi
<azonenberg> gtkmm's C-then-C++ bits are starting to get annoying now
<azonenberg> i really want to override one function but it's deep in gtk's c api and there's no way to hook it
<azonenberg> So instead I have to build a whole new TreeStore
<azonenberg> Which works. it's just a royal pain
<_whitenotifier-3> [scopehal-apps] azonenberg pushed 2 commits to master [+1/-0/±5]
<_whitenotifier-3> [scopehal-apps] azonenberg b3d20a0 - ProtocolAnalyzerWindow: created new custom tree model that avoids the derpy O(n^2) insertion of the stock TreeStore. Fixes #316.
<_whitenotifier-3> [scopehal-apps] azonenberg a55835a - ProtocolTreeModel: implemented erase()
<_whitenotifier-3> [scopehal-apps] azonenberg closed issue #316: Protocol analyzer is extremely slow with a lot of elements -
<azonenberg> Gotta love asymptotic speedups
<azonenberg> Took a couple of hours but it's literally orders of magnitude faster
<azonenberg> something that took over an hour before is now a couple of seconds
<monochroma> :O
<_whitenotifier-3> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-3> [scopehal-apps] azonenberg 4c34e6e - ProtocolTreeModel: added pixbuf support
<azonenberg> So yeah
<azonenberg> There's apparently still major asymptotic speedups possible lol
<azonenberg> Also interesting is that i guess the project i'm working on now is the largest protocol dataset i've ever thrown glscopeclient at
<azonenberg> because i would have hit this issue before had i ever tried to decode this many packets
ericonr has quit [Ping timeout: 265 seconds]
ericonr has joined #scopehal
bvernoux has joined #scopehal
<Degi> huh neat
_whitenotifier-3 has quit [*.net *.split]
_whitenotifier-3 has joined #scopehal
juli9610 has joined #scopehal
StM_ has joined #scopehal
Stephanie has joined #scopehal
monochro1a has joined #scopehal
deltab_ has joined #scopehal
maartenBE has quit [*.net *.split]
Stephie has quit [*.net *.split]
deltab has quit [*.net *.split]
StM has quit [*.net *.split]
monochroma has quit [*.net *.split]
maartenBE has joined #scopehal
<_whitenotifier-3> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±2]
<_whitenotifier-3> [scopehal] azonenberg f0c65d3 - SPIFlashDecoder: added 4-byte address support. Fixed bugs surrounding wait state handling in 1-4-4- Fast Read (0xEB). Fixes #419.
<_whitenotifier-3> [scopehal] azonenberg closed issue #419: SPI flash: add support for 4-byte addressing -
<_whitenotifier-3> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-3> [scopehal] azonenberg 6427cdb - SPIFlashDecoder: fixed incorrect handling of dummy cycles for 0x0B fast read command
deltab_ is now known as deltab
monochro1a is now known as monochroma
<_whitenotifier-3> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±2]
<_whitenotifier-3> [scopehal] azonenberg dde7791 - eSPI: implemented decoding of Put Virtual Wire packets and channel 0 configuration. Merge consecutive get-configuration packets with the same address into a poll transaction. Fixes #420.
<_whitenotifier-3> [scopehal] azonenberg closed issue #420: eSPI: implement "Put Virtual Wire" packet type -
<_whitenotifier-3> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-3> [scopehal] azonenberg 173d489 - eSPI: correctly extend duration of merged packets
<_whitenotifier-3> [scopehal-apps] azonenberg pushed 2 commits to master [+0/-0/±3]
<_whitenotifier-3> [scopehal-apps] azonenberg e244446 - Updated submodules
<_whitenotifier-3> [scopehal-apps] azonenberg fb22573 - WaveformArea: when moving an X axis cursor, ensure that protocol analyzers are notified even if the decode is in a different WaveformArea than the current cursor
Tost has joined #scopehal
<azonenberg> mubes: hey, i just built the latest code and got a warning about the parity bit potentially being uninitialized
<azonenberg> I'm pretty sure its a false positive because it's initialized in STATE_IDLE but please double check
<azonenberg> you might have to add a redundant initializer at the start of the decode when it's declared to make gcc shut up
<d1b2> <mubes> Hmm. I'll check that. What compiler version are you using? I didn't get warnings....odd.
sorear has quit [Read error: Connection reset by peer]
<_whitenotifier-3> [scopehal-apps] azonenberg labeled issue #317: Large font sizes for infobox don't render correctly -
<_whitenotifier-3> [scopehal-apps] azonenberg labeled issue #317: Large font sizes for infobox don't render correctly -
<_whitenotifier-3> [scopehal-apps] azonenberg opened issue #317: Large font sizes for infobox don't render correctly -
<azonenberg> mubes: this was a clean compile from a virgin source tree with gcc 8.3.0-6
<azonenberg> possible newer gcc is smart enough to recognize it as a false positive
<azonenberg> but a redundant initializer won't hurt and i like to kill all warnings
<azonenberg> otherwise you get complacent and ignore them and miss real bugs
sorear has joined #scopehal
Tost has quit [Ping timeout: 252 seconds]
bvernoux has quit [Quit: Leaving]
<Degi> If you have two diff pairs, one with two vias, is it enough to lengthen the other pair by 2 * PCB thickness?
<Degi> (For delay matching)
<Degi> (5 GT/s)
<azonenberg> How tightly do you need to match delay between the pairs?
<azonenberg> generally high speed serdes like that have plenty of fifo space for skew compensation
<Degi> Oh, I meant 1.6 GT/s, the 5 GT/s doesnt need matching
<azonenberg> so a mm or two of pair to pair skew is a non-issue
<Degi> Like a 1.6 GT/s parallel DDR interface...
<Degi> Okay nice
<azonenberg> if it's parallel
<azonenberg> that's a different story because now you have to care about matching them to a common clock
<azonenberg> in that case yeah i'd say 2x pcb is a reasonable first order approximation
<Degi> Hm, one UI should be 187 mm... the FPGA can phase shift the clock to get the clock transition centered on the data
<Degi> Yeah, I guess 2x pcb should be fine, I mean 0.01 UI probably doesn't matter that much when I have 0.5 UI space to each side
<Degi> Btw I got OpenEMS kinda working but the differential impedance calculations are somehow way off...
<azonenberg> Well if its not giving good results it's not too useful, is it? :p
<Degi> I mean I probably configured something wrong... The microstrip coupler example is giving good results
<Degi> I really wish that paraview would work and not complain that a library is missing (despite me recompiling it, it wants but I only have 15, not sure how to fix)
<_whitenotifier-3> [scopehal] mubes opened pull request #422: Fix uninitialised variable warning and prettify SWD line reset ind. -
<d1b2> <mubes> I agree completely, I wouldn't submit code with warnings in it. I compiled with 9.3. It is indeed covered by state=STATE_IDLE as an initialiser, implying that parity=0 is triggered in the switch, but I guess it got missed in the earlier version. PR pushed. Also took the opportunity to fix some graphical goodness in case the length of bits aren't all the same in the idle sequence.
<azonenberg> Great
<azonenberg> will look and merge in a bit
<azonenberg> also your AKL-PT2 is on the way
juli9610 has quit [Quit: Nettalk6 -]
<d1b2> <mubes> ah, sweet, ta. Do I owe you any more for shipping?
<azonenberg> Shipping was a whopping $57
<d1b2> <mubes> bloody heck, they'll sting me another $20 on top of that (20% tax + $10 'handling'). These freebies get expensive 🙂 Just PPed you the differnce.
maartenBE has quit [Ping timeout: 268 seconds]