@topspin That sounds awfully close to heresy. With lots of excuses. Which is typical of heretics.
Posts made by cvi
-
RE: I, ChatGPT
-
RE: In other news today...
@dkf said in In other news today...:
For real? Meat and oil.
In the game ... there it gets you a bigger ammo pouch (or similar) and nothing else.
-
RE: In other news today...
@Arantor Yeah, the grind was bad (and uninteresting). But, it wasn't just that, it was also constantly breaking immersion. I don't remember which game it was, but one of them required you to kill a bunch of whales to upgrade something minor (might have been a coin or ammo pouch). I remember thinking "wtf do we need several whole whales for that for? poor things - this is kinda unnecessary". That's not a line of thinking I often have in games.
Either way, one of the other problems was the map being cluttered with shit by default. Some of it one could turn off, at least temporarily. But, yeah, too busy for my taste, too grindy, and constantly pushing you something premium. Fuck that.
-
RE: In other news today...
@Arantor Odyssey was the last one I tried. I was tempted by Valhalla a few times, but Odyssey put me off it each time. It's something about the AssCreed games that just makes them seem like a chore after a bit. Odyseey was especially bad. (It wasn't the first to do the whole "you need to skin 15 whole elephants to upgrade your coin pouch" thing, but IIRC it was pretty bad on that front.)
-
RE: The absolute state of web storage protocols
@PleegWat said in The absolute state of web storage protocols:
@Gustav Guess it was going to be that or "Of course you hardcode all your translations in the binary."
One binary per language.
-
RE: In other news today...
@Arantor I briefly searched for Ubisoft to remind me what they'd done recently. Saw Far Cry 6, AssCreed Valhalla and Anno 1800.
Clearly they dropped the numbering on AssCreed games, but I'm surprised to see that they've kept the numbering for Anno through 1800 different titles...
-
RE: WTF Bites
@Atazhaia There's an error in there. The correct order is 1, 10, 2, ...
-
RE: Internet of shit
@The_Quiet_One Is it livestreaming the current washing program?
But, yeah, , why is it constantly (except around 14:00) uploading stuff?
Edit: But, also, WhyTF would you connect your washing machine to the internet in the first place?
-
RE: I, ChatGPT
@Atazhaia said in I, ChatGPT:
AIoT, because surely nothing will go wrong from combining AI and IoT...
Poorly secured, internet-connected devices with access to significant compute?
*goes looking for a darker hat*
-
RE: Can I borrow an apology?
@topspin said in Can I borrow an apology?:
Fuck. That doesn't even make sense.
I mean, you could write your rendering code in javascript, like, you know starting with like rasterization basics (lines, fills, polygons, ...) and then building up from there. I'm sure *somebody* out there is doing something like that.
-
RE: Can I borrow an apology?
@topspin said in Can I borrow an apology?:
In the end, I'm just not sure who this is for.
Same.
It won't replace dedicated small vector/matrix libraries either. Those don't gain much benefit from the kind of optimizations you'd see here, and ergonomics of the library are much more important.
I guess, I might use it out of convenience if I need to do some one-off linalg but CBA to pull in a third party library? Except that the new stuff has a learning curve, and I suspect that as soon as perf becomes an issue, it's third party library time again.
-
RE: WTF is happening with Windows 10? And nothing else
@Zerosquare I don't even remember how many restarts there have been now (dual-boot with Linux as the default, so each restart requires me to be there to switch the boot item to Windows). But that problem is fixed. One more restart, though.
-
RE: I, ChatGPT
@boomzilla said in I, ChatGPT:
Personally, I'm severely underwhelmed. Sometimes the autocomplete does something useful. Mostly...it's pretty much garbage.
I've only played it with it in some toy samples. But it seemed to shift the burden from writing code initially to debugging and verifying code. For me, that's the wrong direction. I usually know what I want to write. Verifying and debugging the code is the time-consuming part.
I've also observed some ... rather clueless people. They don't know where to start, so having something spit out something -anything, really- gives them a starting point which they can then desperately ducttape and poke. They never really get to the in depth verification+debugging, so from their perspective it seems like a win. And, if there's one thing we can say about the LLMs, is that they always give you something...
-
RE: I, ChatGPT
@Applied-Mediocrity said in I, ChatGPT:
That it was so obviously shit? Anyone with a pair of eyes should have seen it.
IMO, this is the key point. I don't really mind if somebody uses AI as a tool in their creative process, assuming they produce something of good quality.
But - surprise! - that still requires skilled artists, so you can't just fire the whole bunch of them and replace them with a
n underpaid intern off the streetprompt "engineer". -
RE: WTF is happening with Windows 10? And nothing else
Thought: "I'll quickly run some Windows Updates on the dual-boot Windows install while I get some lunch."
How it's going:
And no lunch.
-
RE: In other news today...
@Zecc said in In other news today...:
Not to mention the Teslas.
"Infested with undesirable bugs" isn't referring to the Tesla software?
-
RE: Guy brings down thousands of npm builds
"Imagine you did an experiment, published a package to NPM and now you want to remove your NPM package. You can't do it if other packages are using it," writes Jossef Harush, Head of Software Supply Chain Security at Checkmarx on the company's blog.
Why are you publishing your experimental stuff to some sort of shared public repo, though?
-
RE: Can I borrow an apology?
@topspin said in Can I borrow an apology?:
. more or less a matrix version of a fused-multiply-add
Haven't watched the talk ... but is it able to do any of the fusing/optimizations you would normally want? (I kinda guess "yes", but...)
And are they really from the "operator overloading considered evil" camp?
It seems like they want to use
std::mdspan
instead of defining their own type(s). Overloading operators onstd::mdspan
would have a bunch of consequences outside ofstd::linalg
. -
RE: Aviation Antipatterns Thread
@LaoC How much do they charge for the extra elbow space?
-
RE: I, ChatGPT
@Gustav said in I, ChatGPT:
read-only instance of ChatGPT
But the instances are read-only already? Based on the above talk, only the context changes, but the context is just ordinary input that's prepended to whatever the user inputs.
-
RE: I, ChatGPT
@Applied-Mediocrity And nobody seems to have a solid plan for fixing it. For SQL, it was always obvious how to fix the problem (mostly something along the lines of "don't be an idiot"). But for the LLMs, there is no distinction between data and instructions. And no obvious way of delimiting safe from unsafe input.
-
RE: I, ChatGPT
Prompt injection talk. Not super advanced, but I haven't bothered peeking under the hood of LLMs. This does briefly mention some stuff (e.g., how they keep track of the context ... they don't, the LLM is just fed a whole bunch of history).
Conclusion: LLMs are freaking dumb. People creating automation plugins for them is -on the other hand- freaking scary.
-
RE: Blakey might blow a gasket or somesuch
@HardwareGeek said in Blakey might blow a gasket or somesuch:
When your coworkers tell you, on your first day on the job, what a company you've just started working for, believe them.
We recent were presented a graph of employee satisfaction from a recent survey. Bar-style plot with the y-axis being split between (essentially) happy/unhappy % of the group in question and the x-axis being subdivided into a few buckets of how long they've been here.
At a first glance, it's not so bad. 50%-50% on generally unhappy vs. generally happy. Then you realize that the x-axis is essentially log(time at employer). Then you realize that all the "happy" answers are in the 0-3 months and 3-9 months (or whatever) range, and all the "unhappy" ones are in the in the 2-4, 4-N and N+ year brackets.
Then you realize the person presenting the data, who is also claiming that things aren't that bad -certainly not from what they've seen- is firmly in the 3-9 month bracket.
Well, statistically speaking, they'll come around soon enough.
-
RE: Hacking News
@TwelveBaud said in Hacking News:
Operation Triangulation:
Crazy exploit. Somebody had posted the flowchart elsewhere here, but that doesn't capture half of it. (Also crazy malware. They probably have more validation than most commercial software.)
-
RE: Hacking News
CCC talk about the polish train "DRM" by the polish hackers. (I think there was a post about the topic earlier.)
Audience question:
Have other train companies reached out to you about dodgy train behaviour?
No comment.
-
RE: Does this circular buffer need a mutex?
@Zerosquare said in Does this circular buffer need a mutex?:
FIFO primitives provided by your [...] language
*laughcry*
Well, there's always the moodycamel concurrent queue.
But, I don't remember there being a great many "standard" concurrent FIFOs out there. Nothing springs into mind for Win32, except if you want to abuse the message pump mechanism for UI stuff. Neither for POSIX.
-
RE: The absolute state of web storage protocols
@Arantor said in The absolute state of web storage protocols:
'if the connection stops you have to reconnect'
What do you need to do to establish a websocket connection? (I could Google, but .) Guess a TCP handshake, probably SSL as well and then some other stuff on top of it?
I guess the question is ... how many round trips?
-
RE: WTF Bites
@dkf When trying to figure out how long an 'age' is, Google led me to the Wikipedia page of Astrological Age. Not the page I was expecting to end up on ever, but one of those is 2150 years. Given I wrote 'ages' (plural), we'd be looking at at least 135605 Ms. If we were to compare things to something Amazon's search and filter functionality, it's clearly more than two ages ahead, so I'd say at least 13560480 Ms.
-
RE: WTF Bites
@Gurth It's a relatively small subset of the items anyway. There were over 200 of them. (I was looking at robotics kits. I'm guessing the ones that show up there are the ones that specify a minimum age, but most don't. I'd say it makes some sense to not show e.g. kits for 9+ year olds with the ones for 15+ year olds, but it's clear that not much thought was put into that selection, so you kinda get what you get. IMO it's still ages ahead of most websites in terms of overall usability...)
-
RE: WTF Bites
I couldn't decide if this should go into the Good or Bad ideas thread.
("Minimum age" for the German-disabled)
-
RE: Can I borrow an apology?
@topspin said in Can I borrow an apology?:
Now they’re proposing to add linalg. I already use a library for that (Eigen) and a BLAS implementation (MKL), I doubt the 3 major stdlib implementers have the manpower to beat what’s already available.
It's fortunate that the graphics proposal ended up dead. I think Linalg is more tractable, but there are some parallels. Curious to see where that one is going. My guess is that one or the other (or all) groups of people will discover that the library wasn't designed with them in mind, if it ever makes it into existance.
-
RE: Things that remind you of WDTWTF members
Response handle_prompt( char const* prompt ) { auto const len = std::strlen( prompt )+1; char* p = new char[len]; std::memcpy( p, prompt, len ); return process( prompt ); }
-
RE: Things that remind you of WDTWTF members
@Applied-Mediocrity You have no proof of that and the real is not imprisoned in a sci-fi looking pod, generating power that helps run an AI poised to take over the world.
-
RE: Things that remind you of WDTWTF members
@boomzilla I think the computers are innocent in this and it'd be unjust to destroy them when the fault lies elsewhere. Clearly, the fault lies with people, so that's what we should be focusing our destruction on.
-
RE: WTF is happening with Windows 11? And nothing else
@loopback0 Help, nothing happens to regedit when I type :wq!. I've been resorting to just turning off the main breaker for the house. Can somebody help?
-
RE: I, ChatGPT
@Zerosquare Making the world more shit is not the same as dooming it. And we already have Facebook, TikTok and smart fridges, so those aren't hypothetical any more.
Edit: I also asked Bing about P(doom). IIRC, it claimed that the artificial experts in AI overestimate that number by a factor 3 or 5 over other people. (It then also claimed that AI isn't dangerous at all and we shouldn't be worried (but please subscribe to the monthly plan), so there's that.)
-
RE: I, ChatGPT
@DogsB P(doom) = 1 if you think long term enough.
AI is competing with a lot of other things on that front. I'm not convinced AI will win.
IMNSHO: Silly Valley Bros thinking that they might be the ones dooming humankind is them vastly overestimating their own importance.
-
RE: WTF Bites
$ time find . -printf "%s\n" | wc -l 50012 real 0m0.452s user 0m0.026s sys 0m0.181s $ time find . -printf "%s\n" | wc -l 50012 real 0m0.096s user 0m0.022s sys 0m0.074s (drop caches again) $ time du -sh 2.6G . real 0m0.420s user 0m0.016s sys 0m0.168s $ time du -sh 2.6G . real 0m0.072s user 0m0.010s sys 0m0.062s
There are about 6k subdirectories and the rest is regular files.
Edit:
@Tsaukpaetra said in WTF Bites:Not if you want to check that everything should fit.
Fair point. If you want to check that things fit, you'd have to at least compute the total size. (Not sure if that's enough when dealing with tons of small files ... depending on how the FS handles those.)
-
RE: WTF Bites
@sebastian-galczynski said in WTF Bites:
scp would be just as slow,
Timing the same thing as @LaoC
$ time find . | wc -l 50012 real 0m0.279s user 0m0.027s sys 0m0.062s $ time find . | wc -l 50012 real 0m0.058s user 0m0.034s sys 0m0.044s
First is after dropping caches, second is cached. Either would never have been enough time to take a screenshot.
Recall that that part didn't even copy any files yet, so whatever Windows was doing there would be strictly over actually transferring stuff.
Even if enumerating files for some reason takes a lot of time, one should be able to start copying as soon as the first file is discovered.
-
RE: In other news today...
@PleegWat said in In other news today...:
Those who do value their sanity can write parser-generators in tools like lex or bison.
How do you rate PEGTL in terms of valuing sanity? It's a BNF-like parser generator, but uses C++ templates.
-
RE: I, ChatGPT
@DogsB said in I, ChatGPT:
I'm doing my best to fill the gap @Gribnit left.
That's a particularly big and strange set of three left shoes to fill.
-
RE: WTF Bites
Dear Windows, I asked you to copy files, not discover them.
Windows "discovering" the files took about as long as scp took to copy them to the network share in the first place. Windows is currently copying files at a rather leisurely pace.
Note to self: next time, just use cygwin+scp on Windows.
-
RE: In other news today...
@DogsB I am actually curious about this too. Not just for Spotify, but for a whole pile of different companies. I wish they'd specify what areas those people are from...
The article mentions a different round with 600 lay-offs total, 200 of them in the "pod casting division". That's a bit more specific, at least for the 200, but I still wonder what those 200 did.
I dunno if this is valid, but from my observations around the recent layoffs, it hasn't affected core engineering staff very much. At least among the people I know, it's definitively been more of the people working on peripheral stuff.
-
RE: WTF Bites
@Tsaukpaetra I'm glad my printer deigned to reply within a few ms, rather than making Windows -and in extension, me- wait.
But you make a valid point. The question is not only "WhyTF do you need to ping my printer if I want to copy a file around to a completely unrelated share?" but also "WhyTF isn't this done in the background, so that I never notice?"
-
RE: WTF Bites
@Watson IIRC, it also varies from printer to printer. Back when we first tested this, we had the brillant idea to embed the EURion pattern into some PDFs that we knew people would try to print. We never got it to work (i.e., printer would refuse to print) on the specific printers at that office.
Edit: The reason I came here is to complain about Windows.
- opens up a network share so I can put a file that I need on the laptop there
- Windows: blocking dialog "Please wait for printer connection [...]"
ARGH. Windows, why do you have to be so ... Windows?
-
RE: Things that remind you of WDTWTF members
@DogsB Crash loop? Isn't restarting your stuff every 10 seconds Ruby-best-practices?
-
RE: WTF is happening with Windows 11? And nothing else
@dkf said in WTF is happening with Windows 11? And nothing else:
But direct framebuffer access has many complications.
These days you wouldn't really need that, most things run through a compositor anyway. So each client gets a private buffer. In X, there's a separate compositor which grabs all the image contents and puts them on screen.
-
RE: WTF Bites
@topspin Hmm.
...
...
New retirement plan. Know how there are those patterns embedded into physical money bills that will make (some) scanners/printers refuse to scan/print them?(*)
Sales idea: Develop software or, perhaps better, hardware that looks for something similar in an image and then edits out the corresponding item. Next, make uniforms/gear for use in e.g. police forces or other paramilitary organizations with that pattern embedded. Talk to government about making that software/hardware mandatory in all future phones, to protect the
childrencitizens or something. (Who needs civil rights and justice when you have them $$$?)(*) I tried it a few times. At one time, a colleague wouldn't believe that this existed, so I took him to the local printerscanner in the office. We put in a €20 of local currency and tried to scan it. Not only did it refuse the scan, but locked down the whole device.
-
RE: WTF is happening with Windows 11? And nothing else
@TimeBandit Sockets are probably better than plain files, because the data can stay in RAM. But there's a lot of overhead in terms of serializing commands and copying data around. A draw call would be something like Application -> Serialize to socket -> Deserialize from socket -> X11 server -> GPU driver -> GPU and so on. You'd be crossing into kernel space twice (once when writing to the socket, once when issuing commands to the GPU driver). The whole thing is rather bad if you need to shift large-ish amounts of data (e.g., stream textures or vertex data).
With direct rendering, you're talking more directly to the GPU driver, something like Application [via OpenGL/Vulkan]-> GPU driver -> GPU and so on. If you're doing things reasonably (e.g., not mega-ancient OpenGL), you ask your graphics API to provide you buffers to load data into, which ideally avoids having to copy that data around more than absolutely necessary (e.g., from RAM to VRAM, though mapping in VRAM into the program's address space can be done). There are also fewer transitions into kernel space (and I believe drivers often try to minimize them further in their user-land components).
-
RE: WTF is happening with Windows 11? And nothing else
@Arantor said in WTF is happening with Windows 11? And nothing else:
The days of having a graphics server in the hardware sense and desktops being kinda thin clients to it is old now, but that was what X was designed for - and I suspect Wayland and friends ignore all that. Even if it has some advantages (e.g. letting you access your machine from another without futzing around with VNC)
X is sort of backwards to what you say. The rendering is done locally, on the machine that you're at. The remote machine, where the application runs, sends drawing commands. If you were doing heavy graphics, you'd want the local machine to be beefy (and not a thin clienty thing). E.g. a silicon graphics workstation or so.
X terminology seems a bit backward in that sense. If you have two machines, local_desktop and remove_server, you run the X server on local_desktop. The X server handles drawing and displaying, so talking to your display, graphics card etc.. If you run an application on remote_server, you can make it connect to the X server. These days, you'd probably tunnel through SSH, but that's not necessary. In this example, the application on remote_server is the client. It sends commands to the server and receives data about events and so on in return.
OpenGL was originally designed towards this. Remote program would send OpenGL commands to your machine. Your machine would execute those commands, potentially on your local GPU. (It still works sort of, but shifting large amounts of data via the network or even just a local socket is not a good idea(tm), which is why stuff like direct rendering exists.)