<_whitenotifier-e>
[scopehal-sigrok-bridge] 602p 40d5793 - Merge master back into flowcontrol; changes to use HzClock
<d1b2>
<louis> I'm still screwing with it. Real diags support helps a lot in debugging this kind of feedback involved system.
<d1b2>
<louis> Have it work reasonably well on the DSCope, but needs some polishing and refactoring until it's something we're going to want to copy around.
<azonenberg>
Ok great keep me posted
<azonenberg>
It already works decently on the picoscope wrt collecting stats
<azonenberg>
you can see we max out the default fifo of 20 waveformss
<azonenberg>
so we get quite a bit of lag
<azonenberg>
From buffer bloat
<azonenberg>
@louis: The other issue i have (which i don't yet have stats for but am working on collecting) is partially acquired waveforms
<azonenberg>
or partially sent i should say
<azonenberg>
so, suppose we are in steady state with the buffer maxed out at 20 waveforms coming from a picoscope
<azonenberg>
right now, there is no explicit backpressure mechanism so the bridge is triggering as fast as it can
<azonenberg>
as a result, it already triggered and acquired a 21st waveform
<azonenberg>
started sending to you, and now send() is blocking because ScopeThread hasn't called AcquireData() yet
<azonenberg>
and the tcp window filled up
<azonenberg>
Now say we stop the acquisition
<azonenberg>
we mark the scope as no longer triggering
<azonenberg>
we flush the fifo
<azonenberg>
(to get rid of stale data when we re-arm the trigger in the future)
<azonenberg>
But that half sent waveform is still in flight
<azonenberg>
nobody will start looking for it until you re-arm the trigger
<azonenberg>
at which point you instantly get an old waveform
<azonenberg>
this is most noticeable if you are doing continuous trigger, stop, then single trigger
<_whitenotifier-e>
[scopehal-apps] azonenberg 512ff2d - Fixed off-by-one in protocol decode rendering causing a large protocol event after a <2 pixel wide one to not be displayed
bvernoux has quit [Quit: Leaving]
bvernoux has joined #scopehal
bvernoux has quit [Quit: Leaving]
<azonenberg>
So I reached out to a prof at UW's EE department about the probing lecture content to see if they were interested in either me giving a guest lecture or adapting the content for them to present in class
<azonenberg>
He said it was probably too advanced for EE 201, the intro lab class he teaches
<azonenberg>
but that he likes the idea and put me in touch with some of the profs who teach more advanced classes
<azonenberg>
So we'll see if anything comes of that