01:23 <azonenberg> also update from UPS, my boards finally left shenzhen

01:23 <azonenberg> probe PCBs are ETA Monday

01:24 <azonenberg> I expect this version to meet the specs i put on the kickstarter but i think i can still do better

02:46 <azonenberg> ok so at this point i have all of scopehal compiling with the new sample format

02:46 <azonenberg> scopeprotocols and scopemeasurements are basically entirely commented out, but what is left compiles

02:46 <azonenberg> working on getting glscopeclient itself to compile now, OscilloscopeWindow has several errors and i suspect other classes (especially WaveformArea) still need a lot of work

02:47 <azonenberg> hoping in the next few hours i'll have a basic #nofilter (just viewing waveforms, no analysis) version on the new architecture ready for testing

04:35 <azonenberg> well, i got it to compile

04:35 <azonenberg> aaaand its segfaults as soon as i actually try to *run* it...

04:35 <azonenberg> but hey, progress :p

04:37 <azonenberg> woo i have a waveform

04:38 <_whitenotifier-c> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±6] https://git.io/Jf0JT

04:38 <_whitenotifier-c> [scopehal-apps] azonenberg b9d6dcb - Refactoring: use new Waveform class

06:44 <_whitenotifier-c> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/Jf0kc

06:44 <_whitenotifier-c> [scopehal-apps] azonenberg 5a5c289 - Improved robustness to malformed input files or missing protocol decodes

10:16 <_whitenotifier-c> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/Jf0mC

10:16 <_whitenotifier-c> [scopehal-apps] azonenberg 217f06f - Disabled OMP in PrepareGeometry as it seemed to make things worse

10:38 <_whitenotifier-c> [scopehal] azonenberg pushed 4 commits to master [+1/-4/±49] https://git.io/Jf0YJ

10:39 <_whitenotifier-c> [scopehal] azonenberg 3228e7f - Refactoring: use new Waveform class. Most decoders disabled since they're not ported to the new object model yet.

10:39 <_whitenotifier-c> [scopehal] azonenberg d8a4981 - Refactored first few filters to use new Waveform class

10:39 <_whitenotifier-c> [scopehal] azonenberg 778a1e4 - DifferenceDecoder: restructured inner loop to be vectorization friendly

10:39 <_whitenotifier-c> [scopehal] azonenberg 51ffa1c - Uncommented debug message

10:39 <azonenberg> ok so at this point the new object model is working but i have ~40 filters not yet ported to it

10:40 <azonenberg> funkylab[m]: so i did a bunch of optimizations and also bumped the waveform size up to 5M points from 1M since there seems to be a fair bit of overhead on the scope for each trigger

10:41 <azonenberg> my most recent test downloaded and processed 92 quad channel waveforms (total 1840M points, 14.7 Gbits) in a minute which comes out to an average rate of 245 Mbps

10:42 <azonenberg> this is pulling waveforms off the scope, subtracting the two diffpairs, 10/100 eth protocol decode on both channels, CDR and eye patterns on both channels, and bathtub plots

10:43 <azonenberg> so in total that's ten filters in the graph

10:44 <azonenberg> and during all of this there's enough idle time that the UI remains very responsive, i can pan around the views with only a little bit of jerkiness

10:45 <azonenberg> this is despite the fact that i'm pushing waveform data to the GPU every frame regardless of whether it's new or not, because some of the conversion from time to screen space happens CPU-side,

10:45 <azonenberg> porting that to the compute shader that does rendering is already on the agenda but it hasnt happened yet

10:46 <azonenberg> at this point i'm pretty happy with the new object model so i'll probably spend most of tomorrow porting the remaining filters to it

10:53 <azonenberg> Deleted a bunch of dead code, removed some overhead, and even vectorized the "subtract" filter. I think there's a lot of room to optimize further wrt special casing sparse/dense sample storage but that's for later

10:53 <azonenberg> the new model has proven its worth so i'll go from there