The Computer And General Tech Thread - Software, Hardware, Questions, etc.

[alice]

I found a neat fast floating-point square root approximation algorithm (which was used in Quake III Arena of all things), and I am going to add it to my code when I get home -- it should make it faster overall.

Fast inverse square root, I think you mean. (It’s pretty famous.)

Ah yes, y = x², inverse of x = sqrt(y). One of the most famous of all functions.

[bradrn]

I found a neat fast floating-point square root approximation algorithm (which was used in Quake III Arena of all things), and I am going to add it to my code when I get home -- it should make it faster overall.

Fast inverse square root, I think you mean. (It’s pretty famous.)

Ah yes, y = x², inverse of x = sqrt(y). One of the most famous of all functions.

Hah!

[Travis B.]

Of course, how you derive a square root from an inverse square root is because sqrt(x) = x * (1 / sqrt(x)).

[jcb]

AI Companies Are Buying Antique Books, Ingesting Their Contents to Train Models, and Then Destroying Them at Incredible Scale

AI companies are literally destroying old and rare books, in an effort to make their slop less sloppy.

[Raphael]

AI Companies Are Buying Antique Books, Ingesting Their Contents to Train Models, and Then Destroying Them at Incredible Scale

AI companies are literally destroying old and rare books, in an effort to make their slop less sloppy.

Can't think of anything else good to say about this other than "To hell with that shit".

[Raphael]

Does anyone know of a good Linux tool, preferably one that runs under Debian or through Flatpak, which can, first, recursively scan a directory with various subdirectories, and then second, tell me if any files have copies in several of those directories? I'm asking because I suspect that some large files might have copies in various places on my computer, needlessly taking up space.

[bradrn]

Does anyone know of a good Linux tool, preferably one that runs under Debian or through Flatpak, which can, first, recursively scan a directory with various subdirectories, and then second, tell me if any files have copies in several of those directories? I'm asking because I suspect that some large files might have copies in various places on my computer, needlessly taking up space.

A quick search turns up an appropriate find incantation:

Code: Select all

find . ! -empty -type f -exec md5sum {} + | sort | uniq -w32 -dD

[Raphael]

Does anyone know of a good Linux tool, preferably one that runs under Debian or through Flatpak, which can, first, recursively scan a directory with various subdirectories, and then second, tell me if any files have copies in several of those directories? I'm asking because I suspect that some large files might have copies in various places on my computer, needlessly taking up space.

A quick search turns up an appropriate find incantation:

Code: Select all

find . ! -empty -type f -exec md5sum {} + | sort | uniq -w32 -dD

Thank you! Currently trying it out.

[Raphael]

Graphic design related question: is there some way to create a graphic that looks as if it would be a pixelated photo, but isn't?

Basically, what I'd like to end up with would be a collection of fairly small squares in many different shades of light pink, so that it ends up looking as if someone would have taken a screencap from a porn video featuring naked white people and then put it through a pixelator.

And before anyone suggests it, no, I have absolutely no intention of getting myself an actual porn video featuring naked white people, taking a screencap from it, and putting it through a pixelator. If I'd want to do that, I wouldn't be asking this question.

I guess I could fire up a graphic software, create a grid of many little squares, and then fill in each and every one with a slightly different shade of pink by hand. But I wonder if there's an easier and faster way to do that.

[bradrn]

I guess I could fire up a graphic software, create a grid of many little squares, and then fill in each and every one with a slightly different shade of pink by hand. But I wonder if there's an easier and faster way to do that.

Well, the ‘easier and faster way’ that comes to mind is to get an actual image and pixelate it. Failing that, if you want pixel art that’s not derivative of any preexisting image, I see no way around creating it by manually colouring in each pixel yourself.

That said, the ‘preexisting image’ need not be from a video. Presumably you could copy-and-paste two or three random pictures of humans, then pixelate that image. If your final resolution is sufficiently low then the result would probably be indistinguishable.

[zompist]

Graphic design related question: is there some way to create a graphic that looks as if it would be a pixelated photo, but isn't?

In Photoshop (or Photoshop Elements, which I use), this is trivial: Filter > Pixelate > Mosaic.

Oh, if you specifically want something that looks like pixelated porn, the easiest way is to pixelate a picture of something that isn't dirty at all but has the right colors. Pictures of elbows or even peaches are good for this.

[bradrn]

Oh, if you specifically want something that looks like pixelated porn, the easiest way is to pixelate a picture of something that isn't dirty at all but has the right colors. Pictures of elbows or even peaches are good for this.

I recall seeing a page several years ago with a collection of images ‘in the wild’ which were rated by an AI as being NSFW despite being completely innocuous on closer inspection. Some of them were really quite impressive in this regard.

(Alas, searching doesn’t seem to turn up the page I remember.)

[Raphael]

Oh, if you specifically want something that looks like pixelated porn, the easiest way is to pixelate a picture of something that isn't dirty at all but has the right colors. Pictures of elbows or even peaches are good for this.

Thank you, the peaches suggestion works really well! I'm about to put the result in the US politics thread.

[Travis B.]

I have implemented globbing for zeptoforth. It enables matching file and directory paths in the current filesystem against glob patterns containing the wildcards *, which matches any number of arbitrary bytes, and ?, which matches any single byte, and passing them to a word specified by an execution token.

Note that . and .. are not matched by wildcards, to avoid unintended matches with unexpected results. It can start from either the current directory or the root directory of the current filesystem based on whether the pattern begins with /. Also, if a pattern ends in / it will match only leaves that are directories; similarly, if the last element of a pattern contains . but is not just . it will match only leaves that are files.

It is a loadable extra but is loaded by the PicoCalc installer scripts for zeptoforth by default except on the rp2040 platform for lack of flash dictionary space.

The code is at https://github.com/tabemann/zeptoforth/ ... on/glob.fs.

[Raphael]

A somewhat old but still apparently fine laptop for whim I'm in charge of family tech support might have become unusable because the charger doesn't seem to wotk any more and might be impossible to replace. Sigh.

[Raphael]

A somewhat old but still apparently fine laptop for whim I'm in charge of family tech support might have become unusable because the charger doesn't seem to wotk any more and might be impossible to replace. Sigh.

Solved!

[Raphael]

In "AI"-related news, here's a bit of a look into the details of how the "AI" companies get the books they turn into "AI" slop:

Irish Times: A ‘bananas’ order for 5,000 obscure titles from a bookshop in Galway fuels suspicion

https://www.irishtimes.com/world/europe ... oksellers/

[Travis B.]

Someone went ahead and implemented being able to mount an RP2350-based board running zeptoforth as a USB mass storage device on a Linux PC (Windows apparently asks the user if they want to format the mass storage device). They showed me their work earlier today. It does not currently work on the RP2040, probably due to alignment issues (the RP2040 does not permit unaligned accesses without extra work to break them up into smaller accesses), which of course can be resolved.

I have been doing some cursory testing of it, and it seems to coexist with accesses from the Forth side surprisingly well. There have been a few oddities, such as when I have updated a file's contents from the Forth side yet Linux insists the file has not changed even though I had not to my knowledge opened the file under Linux since I had mounted it (note though that the file browser used by XFCE may have actually opened the file, causing it to be cached, when scanning for file types, which would explain this behavior). Also, I often have to make sure to execute `sync` because otherwise file operations on the Linux side often are not reflected on the Forth side.

[Raphael]

Congratulations on getting to see people doing interesting things with your stuff!

[Raphael]

Hey, does anyone else aside from me remember the oh-so-informative interactive multimedia CD-ROMs of the 1990s? Those were weird.