<_whitenotifier-f>
[scopehal] hansemro 5855eda - SiglentSCPIOscilloscope: Fix invalid read on :WAVEFORM:WIDTH set commands We do not expect a reply from `:WAVEFORM:WIDTH <BYTE|WORD>` set command. This fixes some minor slowdowns when getting/setting ADC resolution.
<_whitenotifier-f>
[scopehal] hansemro 3813a29 - SiglentSCPIOscilloscope: Fix setting :WAVEFORM:WIDTH with delays This fixes a SDS2000X Plus bug where :WAVEFORM:WIDTH does not change when setting a different ADC mode. Tested bug in firmware versions 1.3.9R10 and 1.5.2R3.
<_whitenotifier-f>
[scopehal] hansemro f14fb82 - SiglentSCPIOscilloscope: no SDS2000XP 10-bit data size workaround for 1.3.9R10+ This resolves read timeouts on SDS2000X Plus running firmware version 1.3.9R10 and newer. Presumably, Siglent fixed the data size issue reported on 1.3.9R6.
<_whitenotifier-f>
[scopehal] azonenberg 579683c - Merge pull request #785 from hansemro/siglent-sds2000xp-10bit-fix [Siglent SDS2000X Plus] Fix ADC mode setting; Fix slow 10-bit data acquisition with newer firmware
<d1b2>
<hansemro> I found some undocumented license commands that are present in an older firmware. LCISL? <OPTION_TYPE> returns either NO_SUPPORT or the license. Don't know if all versions support this, however.
<azonenberg>
oh interesting. May be worth adding it at least in the version that you know it's available int
<d1b2>
<hansemro> LCISL can also be used to install licenses, but that is not really important here. Welcome to the SCPI instrument 'Siglent SDS2204X Plus' >>LCISL? MSO LCISL NO_SUPPORT LCISL MSO,<key> LCISL? MSO LCISL <key>
<d1b2>
<hansemro> The available options are: AWG, WIFI, MSO, FLX, CFD, I2S, 1553, PWA, MANC, SENT
<azonenberg>
File a ticket so we can track it, not a priority to actually implement right now
<_whitenotifier-f>
[scopehal-apps] azonenberg 0f82ffb - Reduced threshold for Filter_FFT sinec it was a bit too aggressive and giving false positive test failures on good-enough matches
<_whitenotifier-f>
[scopehal-apps] azonenberg 2c261ad - Filters: make sure waveforms have valid timescale (now required by some of the ones we have tests for). Print delta and tolerance on failure
<azonenberg>
ok so a few updates: i fixed the failing unit tests
<azonenberg>
picoscope driver now does conversion on the CPU
<azonenberg>
and most importantly, we now support VK_KHR_push_descriptor which means we have the option of submitting the same shader more than once in a given command queue even if the buffers are different
<azonenberg>
(if the GPU supports this, that is)
<azonenberg>
Which should enable some interesting optimizations down the road especially in the rendering/tone mapping path. in general i want to try to use larger command buffers with more work in them whenever possible, and not have to do blocking submits with lots of delays
<azonenberg>
But we'll always need fallback here because we can't guarantee everything supports it
<d1b2>
<johnsel> nice work azonenberg
<azonenberg>
sorry i meant picoscope driver does conversion on the GPU
<azonenberg>
it takes slightly longer than the CPU version in a strict side by side benchmark but i think it's a net performance gain because the next shader using the waveform doesn't have to block on a transfer
<azonenberg>
In general i want to spend some time optimizing how we do dispatches and transfers at some point
<d1b2>
<johnsel> I realized. That is interesting although ofcourse somewhat expected. I keep saying it but I really want XDMA (or RDMA in your case) working
<d1b2>
<johnsel> I unfortunately bought a card that although it can do RDMA it's limited in functionality so that sucks. But there are cheap ebay Mellanoxes that do have the full V2 implementation and support (and can do 25/40, one of the 2) for a few 10s of bucks
<bvernoux>
azonenberg, Do you plan to update roadmap on ngscopeclient ? (that will help to share works on different remaining things ...)