05:04 <azonenberg> hooo boy this is going to be fun

05:04 <azonenberg> chasing disparity errors in 8b10b decoding during pcie decoding

05:04 <azonenberg> what i think is probably happening is that when the link re-negotiates at a higher speed it mis-guesses what the current disparity is

05:05 <azonenberg> So i think what i need to do is maintain a "superposition" disparity state, assuming we could have started out positive OR negative

05:05 <azonenberg> until we end up in a situation in which only one state is possible

07:33 <_whitenotifier> [scopehal] azonenberg pushed 4 commits to master [+0/-0/±9] https://github.com/glscopeclient/scopehal/compare/ca64e22680ec...8f283d72c923

07:33 <_whitenotifier> [scopehal] azonenberg f15f526 - ClockRecoveryFilter: improve auto-sync when squelch opens

07:33 <_whitenotifier> [scopehal] azonenberg 650ed35 - 8b10b: re-sync when there's a gap in the data stream

07:33 <_whitenotifier> [scopehal] azonenberg 0a83a59 - Initial work on decoding electrical idle state. Fixes #387.

07:33 <_whitenotifier> [scopehal] azonenberg 8f283d7 - Lots of progress on PCIe link training decode, can now decode full link startup from power on through 2.5 GT/s initial negotiation to L0 at 5 GT/s. Fixes #638. Fixes #748.

07:33 <_whitenotifier> [scopehal] azonenberg closed issue #387: PCIe decode: handle link power management - https://github.com/glscopeclient/scopehal/issues/387

07:33 <_whitenotifier> [scopehal] azonenberg edited issue #387: PCIe decode: handle link power management - https://github.com/glscopeclient/scopehal/issues/387

07:33 <_whitenotifier> [scopehal] azonenberg closed issue #638: PCIe: decode gen 1/2 link training - https://github.com/glscopeclient/scopehal/issues/638

07:33 <_whitenotifier> [scopehal] azonenberg closed issue #748: PCIe: gen2 speed transition support - https://github.com/glscopeclient/scopehal/issues/748

07:33 <_whitenotifier> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±1] https://github.com/glscopeclient/scopehal-apps/compare/e457cbf60894...d48c7411061f

07:33 <_whitenotifier> [scopehal-apps] azonenberg d48c741 - Updated submodules

08:07 <azonenberg> and no, turns out the actual problem was a bug in the CDR skipping cycles when linking up at a new speed

17:25 <bvernoux> azonenberg, maybe just using FFT

17:26 <d1b2> <david.rysk> How far are we from having a replacement for FFTS?

17:27 <azonenberg> @david.rysk: FFTS has been fully replaced with vkFFT except for a) the spectrogram filter, which is currently #ifdef'd out on non-x86 and b) a couple of unit tests which will probably be replaced with FFTW (since it doesn't matter if test case binaries we never redistribute are GPL contaminated)

17:31 <bvernoux> azonenberg, Any plan to save/restore configuration with data ?

17:31 <bvernoux> azonenberg, in ngscopeclient of course

17:33 <azonenberg> Yes

17:33 <azonenberg> This has been on hold for the last few weeks because we had planned for lain to work on it. She was busy then taking off for the holidays

17:34 <azonenberg> and is hopefully going to start work any day now

17:34 <azonenberg> if i had known it'd take this long i'd probably have started myself but oh well

17:35 <azonenberg> i had plenty of other featurers to work on in the meantime

17:35 <azonenberg> features*