Tootfinder

Mastodon question, I'm trying to completely filter/hide posts with specific hashtags and I can't get it to work.
I set them up as filter for "#blah" and I'm enabling to filter the Home timeline but they just stay in my Home timeline—I've tried with and without "whole word" on.
Is there some known issue in v4.5.7 or am I missing something?

Sieg für YouTube: Warum Plattformen nicht für jeden Upload haften
Ein US-Urteil bestätigt, dass YouTube & Co. nicht proaktiv nach Urheberrechtsverletzungen suchen müssen. Das gilt auch, wenn sie fortschrittliche Filter nutzen.

Filter in Onlineshops: Für Nutzer praktisch, für SEO oft eine Katastrophe. 💥
Jede Filterkombination (Farbe Größe Preis) erzeugt eine eigene URL. Bei großen Shops entstehen Millionen Seiten mit fast identischem Inhalt.
Google verbrennt sein Crawl-Budget in diesem URL-Labyrinth statt die wichtigen Kategorien zu indexieren. Die Seiten die ranken sollen, werden nicht richtig gecrawlt.
Lösung: Filter-URLs per Canonical oder noindex auf die Hauptkategorie lenken. 🏪

Added a scrollable list of errors to the filter graph editor so you can see multiple problems more clearly.
The error message automatically spawns, docked to the south side of the filter graph editor, when an error appears and disappears when the last error is resolved.

I can’t believe I’m having to kludge a CSS filter onto their iframe to invert the embedded checkout form of a company that carries out *trillions of dollars* of transactions every year because they don’t have dark mode support.
My feeble mind clearly cannot understand such high capitalist craft.

»YouTube will gegen KI-Müll vorgehen, aber KI-Influencer fördern:
Bessere Filter sollen dafür sorgen, dass nur "hochwertige" Inhalte bei den Zuseherinnen und Zusehern ankommen.«
Das hochwertig in Anführungszeichen steht mMn ganz richtig. Schade dass @… im Fediverse noch nicht so populär ist und leider auch dort einiges umstrittenes veröffentl…

#PhanpySocial changelog ✨
📤 Allow receiving shared data with the Web Share Target API
🔐 Timeline access controls
👤 New shortcut: "Profile"
📝 "Only followings" filter for Mentions
↕️ Preliminary support for vertical-lr writing mode
🐛 Bug fixes
🔗

Luckily, I can use last year's lecture notes unchanged for the course that's about to start in [checks watch] 10 minutes, because that's not what it's supposed to look like…
I’m trying to use the #Pandoc diagram.lua filter with mermaid to produce PDF via #Typst. Apparently the p…

Luckily, I can use last year's lecture notes unchanged for the course that's about to start in [checks watch] 10 minutes, because that's not what it's supposed to look like…
I’m trying to use the #Pandoc diagram.lua filter with mermaid to produce PDF via #Typst. Apparently the p…

Luckily, I can use last year's lecture notes unchanged for the course that's about to start in [checks watch] 10 minutes, because that's not what it's supposed to look like…
I’m trying to use the #Pandoc diagram.lua filter with mermaid to produce PDF via #Typst. Apparently the p…

RED （Ocassionally） II 🔴
红 （有时） II 🔴
📷 Zeiss IKON Super Ikonta 533/16
🎞️ Harman Red 125 (6x6)
#filmphotography #Photography #blackandwhite

New record scopehal filter speedup: Vector Frequency, which computes the instantaneous frequency of an i/q signal as d/dt(atan2(q,i)).
The (single threaded, non simd) CPU version took about 1.5 seconds to crunch 50M samples on my Xeon 6144.
The trivially simple GPU version? About 1.5 *milli* seconds on a 2080 Ti.

Landgericht Berlin: Doctolib-Filter für Kassenpatienten irreführend
Doctolib darf beim Filter „Nur Termine mit gesetzlicher Versicherung anzeigen“ keine Termine von Privatpraxen einblenden.
https://www.…

Neat ideas for navigating threaded comments & new comments highlighting. I kinda thought of this but really wonder about its usefulness.
- If there's too much nesting of comments, the sidebar space won't be enough
- It shows "[INT] [NAME]" where INT = vote count. Higher count is probably more read-worthy?
- Per-day filter for new comments highlighting is probably too broad in scope if the conversation is too active.
- Sidebar won't appear on sma…

Filter in Onlineshops: Für Nutzer praktisch, für SEO oft eine Katastrophe. 💥
Jede Filterkombination (Farbe Größe Preis) erzeugt eine eigene URL. Bei großen Shops entstehen Millionen Seiten mit fast identischem Inhalt.
Google verbrennt sein Crawl-Budget in diesem URL-Labyrinth statt die wichtigen Kategorien zu indexieren. Die Seiten die ranken sollen, werden nicht richtig gecrawlt.
Lösung: Filter-URLs per Canonical oder noindex auf die Hauptkategorie lenken. 🏪

"Fish biology inspires laundry microplastics clean up"
#Microplastics #Laundry
https://www.

🇺🇦 #NowPlaying on KEXP's #WoPop
Yuki Kawamura:
🎵 Magic feat. Color Filter
#YukiKawamura
https://oiranmusic.bandcamp.com/track/magic-ft-color-filter

The calendar service I use (Infomaniak's kSuite) started complaining that some calendars I import as ICS URLs (Google calendars of relatives and associations) were "too big".
It refuses to synchronize calendars that contain more than 25,000 events.
So I created a small proxy, to filter the oldest events, and now my calendar and I are happy.
For any calendar https://original-calendar/file.ics, I now give it:
https://my-service/ics-filter?url=https://origina…

We really need a new Internet. How do we help?
https://torrentfreak.com/italy-fines-cloudflare-e14-million-for-refusing-to-filter-pirate-sites-on-public-1-1-1-1-dns/

Dat mooie weerbericht van in mijn geval @… geeft ook de luchtkwaliteit weer. Dan is het zeer plezierig om te zien dat die binnenshuis beter is dan buiten en dat de gecombineerde filters van de balansventilatie en de extra investering in het BrinkPureInduct filter de PM 2.5 meestal op nul kunnen krijgen.

Es ist voll okay, wenn ihr zur aktuellen Kommunalwahl postet. Aber bitte macht doch einen Hashtag #politik oder #kommunalwahl dran.
(und wahrscheinlich sehe ich dank Filter gleich meinen eigenen toot nicht mehr 🤔).

iPhone-Tool iMazing: Version 3.5 mit Dateisuche, Messaging-Filter und mehr
DigiDNA aus der Schweiz hat iMazing aktualisiert. Das Update bringt eine Reihe neuer Features, auf die Nutzer länger gewartet hatten.

Summary of the recent optimization work in ngscopeclient (not counting core/driver stuff, just filters)
I'm now making a pass through the entire filter library in alphabetical order removing some deprecated APIs, adding error reporting, and doing trivial GPU optimizations where I find them but not spending more than a few hours on a filter unless it's something like the CDR or eye that is really critical.
None of these are what I would consider "done". I'm f…

FTW. You search for "coffee filter" and all 1000 results are for ground coffee to use in, and you won't ever guess this, *coffee filters*. 🤦‍♂️
#stupid #technology #ecommerce

OTTL context inference comes to the #Filter Processor
https://opentelemetry.io/blog/2026/ottl-context-inference-come-to-filterprocessor/

#Steady-#Klimacrew
Im #Nachrichtengeschäft nimmt die Menge an medialen Inhalten exponentiell zu, und damit auch die Häufigkeit potenziell verstörender und traumatisierender

Why don’t the LLM harnesses that support MCP slap jq in front of the MCP response? Some of the MCPs are *very* verbose (looking at you Jira), and allowing the model to filter the response before it enters the context window would be very useful

Design and Evaluation of a GPU-Accelerated Cuckoo Filter
#cuckoofilter #computerscience
🔗 Thesis https://tdo…

@… @… same, even have a unittesting-filter target for this

The Accessibility Conformance Testing (ACT) Task Force has published its rules, letting you filter automated rules:
https://www.w3.org/WAI/standards-guidelines/act/rules/?requirements=a,aa&status=approved&imp…

I've been working on assembling my one-kilogram QSO¹ kit. I set a 1kg limit including the case into which to cram everything I need to work POTA on a whim because it's just bouncing around in the bottom of my backpack. I'm still working out what to carry for filter, choke, and ununs; after I'm done, I'll write up the whole kit with pictures.
I just swapped out a 20Ah Anker power bank for an 88g

A TV (470MHz-790MHz) filter arrived today to try to cope with the temporary traffic light problems I mentioned the other day; they took the traffic lights away yesterday. Oh well, useful to have in the box for next time there's a concert on.

EDIT: solved, see answers
--
sorry to bother you, but I'm wondering:
Is there a better way to find past conversations with another user?
I mean written interactions in posts, or the so-called DMs.
Maybe a filter reply notifications by user or date? Or a search for username word?
I can't find some info a mutual gave me, while we were commenting a post I can't remember, and my only idea is going through all my reply notifications
Thanks!
#AskFedi

Aaaand this is why I have 40GbE to my desk.
Early test, I need a bunch of optimization to do (multiscope is not as performant as I want I suspect there's mutex issues and/or unnecessary filter graph refreshes).
But here's two thunderscopes hosted on separate machines, each with 10GbE to the core switch, then 40G from there to my workstation.

Largest federal employee union calls for Noem, Miller's resignations (ABC News)
https://abcnews.go.com/US/live-updates/minneapolis-ice-shooting-live-updates-doj-investigating-apparent/?id=129340693&entryId=129600588&userab=abcn_du_cat_topic_feature_holdout-474*variant_b_redesign-1939,abcn_popular_reads_exp-497*variant_c_3days_filter-2078,abcn_ad_cadence-481*control-a-1962,abcn_news_for_you_exp-496*variant_a_control-2074,otv_web_content_rec-445*variant_c_trending-1851&userab=abcn_du_cat_topic_feature_holdout-474*variant_b_redesign-1939,abcn_popular_reads_exp-497*variant_b_7days_filter-2077,abcn_ad_cadence-481*control-a-1962,abcn_news_for_you_exp-496*variant_a_control-2074,otv_web_content_rec-445*variant_b_less_popular_bias-1850
http://www.memeorandum.com/260127/p142#a260127p142

Any #ffmpeg experts here? I want to do one simple thing: make certain colors transparent (then alphaextract to feed into tesseract #OCR). That's it. The colorkey filter is perfect, but it seems to be an absolute impossibility if you want more than one color transparent.

"Slop filter" ->
Nvidia, This Looks Terrible... - SomeOrdinaryGamers
https://www.youtube.com/watch?v=kYgfX_uy8G8

Incroyable, contrairement aux autres je suis super excité par cette technologie. C'est un bond en avant énorme. C'est pas un beauty filter, toutes les textures, les lumières, l'ambiance est bien plus léchée avec cette techno.
J'ai hate de pouvoir tester !
Il ne faut pas se dire que c'est pas exactement le look d'avant 5.0, il faut regarder l'image telle qu'elle est.

A Novel Explicit Filter for the Approximate Deconvolution in Large-Eddy Simulation on General Unstructured Grids: A posteriori tests on highly stretched grids
Mohammad Bagher Molaei, Ehsan Amani, Morteza Ghorbani
https://arxiv.org/abs/2602.21166 https://arxiv.org/pdf/2602.21166 https://arxiv.org/html/2602.21166
arXiv:2602.21166v1 Announce Type: new
Abstract: Explicit filters play a pivotal role in the scale separation and numerical stability of advanced Large Eddy Simulation (LES) closures, such as dynamic eddy-viscosity or Approximate Deconvolution (AD) methods. In the present study, it is demonstrated that the performance of commonly used explicit filters applicable to general unstructured grids highly depends on the grid configuration, specifically the cell aspect ratio, which can result in poor filter spectral properties, ultimately leading to large errors and even solution divergence. This study introduces a novel, efficient explicit filter for general unstructured grids, addressing this shortcoming through a combination of a face-averaging technique and recursive filtering. The filter parameters are then determined through a constrained multi-objective optimization, ensuring desirable spectral properties, including high-wavenumber attenuation, filter-width precision, filter stability and positivity, and minimized dispersion and commutation errors. The AD-LES of turbulent channel flow benchmarks using the new filter demonstrate a noticeable improvement in turbulent flow predictions on highly stretched boundary-layer-type grids, particularly in reducing the log-layer mean velocity profile mismatch, compared to simulations using conventional filters. The analyses show that this enhancement is mainly attributed to the sufficient level of attenuation near the Nyquist wavenumber achieved by the new filter in all spatial directions across various grid configurations, among others. The new filter was also successfully tested on unstructured prism grids for the 3D Taylor-Green vortex benchmark.
toXiv_bot_toot

#Blakes7 A07 - Mission To Destiny
KENDALL: Why? For what reason?
CALLY: Perhaps Sonheim was still trying to cover up. Mandrian knew about him and threatened to expose him. Or perhaps he went into the filter plant while Sonheim was sabotaging the equipment?

{dtrack} makes documentation of data wrangling part of the analysis and creates pretty flow charts: #rstats

"Washing machine filter captures microfibers as small as 20 micrometers in size"
#Plastic #Plastics
https://

Nice, a uBlockOrigin blocklist for AI-based sites:
#AI #LLM

Sentry's strategy is brilliant: either you pay to increase log volume or you pay to filter logs so you don't have to pay to increase log volume.
Yeah yeah I know you can filter out at the source but you know what I mean.

Google hat FID als Web Vital durch INP ersetzt. Klingt nach Buchstabensuppe, ist aber wichtig. 🍜
INP misst nicht nur den ersten Klick, sondern jede Interaktion auf der Seite. Menü öffnen, Filter setzen, Accordion aufklappen, Tabs wechseln.
Viele Seiten hatten perfekte FID-Werte, aber Nutzer haben trotzdem bei jeder Aktion gewartet. INP deckt genau das auf.
Wenn die Seite beim zweiten Klick hängt, hilft der schnelle erste Eindruck nicht. DevTools zeigt INP live. ⏳

Regarding that DF DLSS 5 preview, I think it needs to be said how the untreated RE9 footage looks more realistic than the filter aiming at realism, maybe focus less on trying to just give us a way less efficient version of Reshade and just make optimizing for the ACTUAL artistic intent easier.

Ein neues Filtersystem nach dem Vorbild von #Fischmäulern kann über 99 % des #Mikroplastiks aus dem Abwasser von #Waschmaschinen entfernen.
Forschende aus

Some _very_ early algorithm concept/pre-viz sketches/explorations for the Tron Legacy (2010) intro sequence. They informed another concept/approach in which we applied a similar algorithm to progressively trace out edges of an initially invisible and super detailed 3D city mesh. This required a lot of effort to retopologize the (huge) geometry supplied and creating a multi-res navigation graph to prioritize long/major edges over shorter ones, filter out undesired edges/directions, thereby cr…

Had to add a filter for all these "verification" spambots...

Hey folks, please, if you are posting especially political stuff, photos, articles, memes... Please use a common hashtag.
If someone can't filter away based on hashtags, the only solution is to mute the whole account. This would be collateral damage.
Cant speak for anyone else, but I want to keep my timeline clean. Even if I have to mute complete accounts 🙁

Y-Y-Youw UwU skiww has wevewed up?!?1
(The UwUifier now translates in both directions... kinda)
https://uwuifier.com/?language=UWU_TO_ORG

A look at iOS 26's call-screening feature, which some say is like having an assistant to filter out unwanted calls, as some complain it makes them miss calls (Andrew Zucker/Wall Street Journal)
https://www.wsj.com/style/iphone-call-scre

Started collecting some performance data on ngscopeclient filters as I go through and optimize/refactor.
This is nowhere near the entire filter suite, and a few rows are missing data right now.
8B/10B is still running on the CPU, as you can probably guess from the abysmal throughput, but I had it in one of my test filter graphs so I included the data.

I guess it's time to filter out any mentions of Bandcamp for today if I want to eat anything this month.

Wer definiert #Desinformation - wer entscheidet was #Wahrheit ist?
Könnt ihr euch noch an die Argumente gegen die #Uploadfilter aus Artikel13/17 erinnern? 🤔

I don't suppose there's a filter to transform a #Convo site into some sort of #WordPress theme? Or a skilled hand who's done it before?

I’m happy to report that @…’s multibib Pandoc Lua filter works great for this purpose: you sort your references into different bib files according to your needs (books, journal articles, conference papers, etc.), indicate in the document where the sections are to appear, and use the nocite metadata field to fill them.
If you use Typst as backend, and if you al…

I’m happy to report that @…’s multibib Pandoc Lua filter works great for this purpose: you sort your references into different bib files according to your needs (books, journal articles, conference papers, etc.), indicate in the document where the sections are to appear, and use the nocite metadata field to fill them.
If you use Typst as backend, and if you al…

Cat Flowbee! @… https://infosec.exchange/@atax1a/115839544124655751

A few months after handing in my thesis and receiving my Master of Science degree in 2022, ChatGPT was released. Essentially, nothing about how I solve problems has changed since then, but I had to continuously observe the brain drain around me. It's weird how the majority of people suddenly treat anything an LLM spits out as truth while I stick to good old search engines (with the additional hurdle of now having to filter LLM garbage from my search results).

I saw a video where a guy built a Power BI app to filter only very specific senders for a client who got a lot of spam email. I commented, "there's a checkbox in OWA settings to only deliver emails from your safe senders list." Could've done it in like 2 clicks.

@… Det finns en bra början i den här nätverkskartan över makthavare i skolan av Magasinet Filter som Tankesmedjan Balans hänvisar till här: https://tankesmedjanbalans.se/balans-kommenterar-podavsnittet-zombieskolan-john-bauer/
…

As usual, I just look at somewhere in ngscopeclient and more massive speedups fall out.
Today it's the AC couple filter (10x speedup), average (5.6x), and base (2.4x so far from trivial block swaps, but the inner loop is still on the CPU so I expect a huge speedup once I get *that* in a shader).
I can't imagine how fast things are going to get once I have a sizeable fraction of the filter library GPU accelerated.

The James #Webb Space Telescope would probably not be your first choice for observing the large open #StarCluster of the #Pleiades but the authors of the paper "JWST imaging of the Pleiades: anisotropy of turbulence in the cold neutral medium" - https://arxiv.org/abs/2602.03672 - did it anyway, to study the turbulent energy cascade in the CNM the cluster happens to move through right now. Here is what they saw, with the NIRCam and the F335M filter - for the four cuts from the JWST pictures at the bottom in detail see fig. 2 of the paper.

RE: #PhanpySocialDev

Took my car to the dealer for a complimentary annual maintence today.
They told me the cabin air filter was not covered and it needed to be replaced, they did show me a very dirty filter before asking if I wanted them to replace it.
I agreed in the moment to the $117 cost. I think it was about 50% parts and 50% labor.
But I've now looked it up, and it's an incredibly simple thing to replace and is a $20 part. So never again.
#carRepair #cardealer

Working on improving error reporting in ngscopeclient rather than having filters simply silently produce no output, or log to stdout, when something goes wrong.
Here's my current WIP: the block that has problems displays a warning icon and summary title up top, then you can mouse over it to see more detailed error information.
Thoughts? Anything I should change before I start working on adding this sort of feedback to more filter blocks?

I hadn’t planned to do this today, but I just added a new filter to pandoc-lecturenotes https://github.com/mxpiotrowski/pandoc-lecturenotes, which allows you to produce slides in any of the formats supported by

Large eddy simulation of turbulent swirl-stabilized flames using the front propagation formulation: impact of the resolved flame thickness
Ruochen Guo, Yunde Su, Yuewen Jiang
https://arxiv.org/abs/2602.21940 https://arxiv.org/pdf/2602.21940 https://arxiv.org/html/2602.21940
arXiv:2602.21940v1 Announce Type: new
Abstract: This work extends the front propagation formulation (FPF) combustion model to large eddy simulation (LES) of swirl-stabilized turbulent premixed flames and investigates the effects of resolved flame thickness on the predicted flame dynamics. The FPF method is designed to mitigate the spurious propagation of under-resolved flames while preserving the reaction characteristics of filtered flame fronts. In this study, the model is extended to account for non-adiabatic effects and is coupled with an improved sub-filter flame speed estimation that resolves the inconsistency arising from heat-release effects on local sub-filter turbulence. The performance of the extended FPF method is validated by LES of the TECFLAM swirl-stabilized burner, where the results agree well with experimental measurements. The simulations reveal that the stretching of vortical structures in the outer shear layer leads to the formation of trapped flame pockets, which are identified as the physical mechanism responsible for the secondary temperature peaks observed in the experiment. The prediction of this phenomenon is shown to be strongly dependent on the resolved flame thickness, when the filter size is used for modeling sub-filter flame wrinklings. Without proper modeling of the chemical steepening effects, the thickness of the resolved flame brush is over-predicted, causing the flame consumption rate to be under-estimated. Consequently, the flame brush detaches from the outer shear layer, resulting in a failure to capture the flame pockets and the associated secondary temperature peaks.
toXiv_bot_toot

But it would be nice to have not just a link in the lecture notes, but a preview. And you know what? It’s almost too easy to write a little #Pandoc Lua filter for it ;-)
□
(I intend to write a blog post about it next week or so.)

But it would be nice to have not just a link in the lecture notes, but a preview. And you know what? It’s almost too easy to write a little #Pandoc Lua filter for it ;-)
□
(I intend to write a blog post about it next week or so.)

I've been trying out @… and I really need there to be some way to filter out AI videos. Like, I just *do not want* to see any videos that are AI generated.
Seeing them is reason enough for me to not use Loops.
#loops

I can hardly express how much I love the filter functionality in Mastodon to curate my feed.

The outcome of a bunch of shader tuning last night: the upsample filter (4x sin(x/x) from 20M to 80M points in this test) went from 6.55 ms to 1.5 ms.
Original: 8% of peak DRAM read BW, 31% write, 14% L2$ hit rate.
New (just changed memory access patterns to be more coalesce/cache friendly): 9% read, 37% write, 73% L2$ hit
A similar memory ordering optimization cut the PAM edge detector from about 14 to 10 ms but my SM occupancy is still crap (around 12% of warp slots used)…

🇺🇦 #NowPlaying on #KEXP's #MorningShow
Filter:
🎵 Hey Man Nice Shot
#Filter
https://aggro1.bandcamp.com/track/filter-hey-man-nice-shot-x-limbo-slice-aphotic
https://open.spotify.com/track/1YFNBJbmk9xLlYHXjhK7qz

"Plagued by Flooding, an African City Reengineers Its Wetlands"
#Africa #Wetlands #Environment

So, tonight's goal is to continue with ngscopeclient performance work.
I started out by doubling the speed of the eye pattern *again* by moving index buffer calculation from CPU to GPU.
Next up is going to be getting the 100baseTX decoder to not be so slow. Right now of the 43 seconds of CPU time in the current 1-minute benchmark, 26.9 is spent sampling the MLT-3 waveform on rising edges of the recovered clock.
The thing is, we already *know* the sample values at the re…

I hadn’t planned to do this today, but I just added a new filter to pandoc-lecturenotes https://github.com/mxpiotrowski/pandoc-lecturenotes, which allows you to produce slides in any of the formats supported by

I hadn’t planned to do this today, but I just added a new filter to pandoc-lecturenotes https://github.com/mxpiotrowski/pandoc-lecturenotes, which allows you to produce slides in any of the formats supported by

After a lot of other poking and prodding at other stuff failed, the PAM edge detector got a 30% speedup and the sin(x)/x upsampling filter is about 4x faster just by restructuring some stuff to change memory access patterns.
As a reminder: a GPU is a device for turning compute bound problems into memory bandwidth bound problems.

Thinking about future release plans for ngscopeclient.
There have been massive performance improvements and some significant bug fixes since v0.1.1 and the ThunderScope dev edition units are going to manufacture so we'll be getting a lot of additional users in ~2 months.
But there's also a lot of ongoing backend refactoring and changes to filter graph blocks that won't be strictly backwards compatible (old filter graphs will need updating) and while I freely break sof…

Another evening of filter refactoring and optimization, a few more nice performance jumps.
I've now done a first pass (remove deprecated method signatures, add explicit input location, add NVTX trace data, do easy GPU optimization if I see an obvious low effort win) on all filters A-F alphabetically plus a few later on that were priorities for one reason or other.
90 down (of which 23 were optimized and the rest just refactored), 115 to go.
Some of the remaining ones sh…

"Older petrol and diesel vehicles produce 50% of harmful particle emissions in Finland, study shows"
#Finland #Cars #Vehicles

1.7x speedup on the 100baseT1 decode from doing the PAM3 demodulation (but not the descrambling) in a shader.
Now the whole baseT1 demo is running at about 8 WFM/s, with a roughly 125ms filter graph execution time (on 80ms of waveform data, so real time would be 12.5 WFM/s)
This breaks down (roughly, not exact due to parallel execution) to:
* 1.3 ms subtract diffpair legs
* 10.6 ms de-embed directional coupler
* 0.6 ms equalization
* 7.5 ms upsample
* 14…

Finally had time to sit down and do a GPU version of the constellation diagram filter.
I'm now running this filter graph (4 channels -> 2 differential legs, S-parameter de-embed of dual directional coupler, 3 dB FIR equalizer, 4x sin(x)/x upsample, PAM-3 edge detection, CDR PLL, PAM-3 eye pattern, demux to 2D-PAM3 channels, 2D-PAM3 constellation, 100baseT1 single pair ethernet protocol decode) at about 6.3 WFM/s on 4 channels * 20M points.
This is 504 Msps or a touch over …

Last evening of Christmas vacation before getting back to work for the new year.
Aiming to make it a good one. Working on GPU acceleration of the TIE filter and maybe the histogram block so I can do jitter analysis faster.
The peak at T=0 in the jitter histogram is related to use of the GPU CDR block and doesn't show up in the CPU implementation. It's likely caused by transients of some sort at thread block boundaries and I need to spend more time chasing it, I wouldn'…

"New filtration technology could be gamechanger in removal of Pfas ‘forever chemicals’"
#PFAS #ForeverChemicals

Buttbuttination (n): an assassination carried out in the presence of an overly aggressive profanity filter

I'm going to have to screen record this, I can't believe I got it working this well.
2x 50M point differential Ethernet waveform into subtract filter, CDR, and eye pattern.
Refreshing at 8.3 Hz. With just a little bit more optimization or faster hardware this will be real time.
Then I can start working on getting protocol decodes to run at full rate too.

Now that I'm on a GPU acceleration kick in ngscopeclient again I'm spoiled. Every time I use a filter block even on a massive waveform I expect it to be instantaneous and get frustrated when it's not lol.
If this keeps up we'll have GPU i2c and spi decodes pretty soon.

Initial experiments on a GPU-accelerated parallel CDR PLL filter.
Fundamentally, the problem is that a PLL is stateful so you can't process any given iteration of it without knowing the previous state. I'm trying to work around that by recognizing that the impulse response of the PLL loop filter tails off to effectively zero after a while, so we can truncate and samples older than that point will not materially affect the output.
What you see here is the first pass of wha…

Starting to look at some performance improvements for various scopehal filters as I begin the push for v0.2.
The benchmark dataset is P/N legs of 100baseTX Ethernet sampled at 500 Msps with a ThunderScope.
At 50M point memory depth, each waveform is 100 MB in size (2 channels * 8 bits * 500 Msps) and covers 100ms of real time.
The filter graph currently takes just shy of a second to run, meaning we're at 10% of real time. I want to improve on this.

Looking to have someone redraw the icon for the "coupler de-embed" filter in ngscopeclient as I'm not thrilled with the current one.
The big question is, what should it actually depict?
The filter performs directivity enhancement for a dual directional coupler. It takes in the measured forward and reverse waveforms, S-parameters of the coupler, and outputs new forward and reverse waveforms with (most of) the leakage removed.

Current summary of all of the GPU acceleration work I've been doing in ngscopeclient since v0.1.1.
Other than the CDR PLL and eye pattern which were high-priority targets that were worth the effort, and the Ethernet which I got nerdsniped on, these are all trivial inner loops of just a few lines that I threw at the GPU without any significant tuning, so there's probably more performance upside.
But I'm not complaining with what I've pulled off so far. (As you ca…

Just an average day of oscilloscope development. If your scope isn't saturating your GPU are you even trying?
Added NVTX support to libscopehal and ngscopeclient so I can have specific filter graph blocks and processing stages show up in profiling traces. Makes things a lot more understandable.

ngscopeclient is pretty fast, but it's never fast enough.
NVTX instrumentation has been super handy for helping to understand what's going on and which shaders are bogging things down.
I really need to make the CDR PLL both faster and not have artifacts at block boundaries. That's my single biggest pain point at the moment.
I'd love to find more people to help out with this side of things eventually. But where the heck do I find an AAA game shader developer …

Debating removing some of the AVX accelerated filter implementations in libscopehal for blocks that we now have GPU support versions of, leaving only the shader and the unoptimized CPU fallback for non-x86 platforms that also don't have a suitable GPU.
For starters, the eye pattern: what are the odds that we have any reasonable number of users where a) they have a CPU with AVX512F (or even AVX2), and b) do NOT have a GPU with GL_ARB_gpu_shader_int64 and GL_EXT_shader_atomic_int64?

Just fixed an interesting and subtle shader performance bug.
Here I'm trying to recover a clock from a PAM-3 signal which consists of two consecutive filter blocks: the PAM edge detector (find level crossings and interpolate, accounting for the fact that the threshold changes depending on start/end symbol), and then the CDR PLL proper.
If you're not familiar with NSight Systems especially on complex multithreaded applications, there's a lot going on here even though I…

Took a look at the ngscopeclient "fall time" filter since it's the next in line alphabetically for some refactoring and decided hey, the inner loop is pretty simple let's try GPUing it.
But first I wanted to get a baseline run time for 50M points (790 ms).
Aaand found numerical stability issues. So I need to fix the algorithm before I optimize it.
This is why we use int64's for time values in all new code, not float32 or even (as done here) float64.