284 MB disc image stored in an 8MB .gz

>284 MB disc image stored in an 8MB .gz
How?

>30GB game compressed to 3GB
Compression algorithms are weird.

>what is file compression

probably lots of zeros.

Some data formats are more (losslessly) compressible than others.
Plain text is very compressible becaues there are lots of patterns that compression aslgorithsm can optimize away.

>there are lots of patterns that compression aslgorithsm can optimize away.
How do they do this?

I could store one million zeros as 1,000,000 x "0" for a crude example.

>not understanding what the word "compress" means

Why not 1,000,000 รท 0 ?

No real reason. That would just be stupid.

Bad things happen when you divide by zero. Stop it.

Information theory.
The minimum size of a data is it's entropy (in bits). If the data is all '0' and sometimes '1' (low entropy), you can use an extended Huffman coding to convert a source material X into Y uniformly distributed (max entropy) containing the exact same data (lossless) just a bit longer than entropy(source) (in bits).

tl;dr, compute the entropy/size of a data to check if it could be compressed.

Because unless there's a lot of images/videos (which are already compressed and thus cannot be compressed even more) most software will happily squash to 1/8th to 1/16th of its size.

or lots of repeating bytes

PPmD is amazing at compressing text
always choose it over lzma2 if you're handling text-only

What are Sup Forums's favorite compression methods? For anything.
I obsessively compress images, sadly my machine isn't fast enough to do it efficiently. Get's to be a pain in the ass.

You can make a 0-byte file any size and then compress it to almost nothing

MB disc image stored in an 8MB .gz

When you compress something, you trade processing power for free space. For example, if you were taking a statistics course, could have a table of values up to 3 significant figurss for a normal distribution. That's a lot of data to store, but quick to lookup.

Or you could calculate it every time you wanted to know a value. It would take 5 times as long, but you wouldn't need to carry a page or 2 of tables with numbers all over them.

every simplify an equation in math class?
basically that

Like this.

I believe that is the question being asked yes

That's nothing.
Check out deduplication.

That's nothing.
You should check out depopulation.
6 million Jews compressed into only a few thousand remains

Spot patterns, find the longest + most used pattern and replace it with 1 then 0 inbetween patterns and 11 for the next most often partern etc
This is like 5 minute intro into compression. but the longer more often used patterns are, the more easily they can be compressed. If you write an essay, that has 1500 words. There are probably only 150-250 unique words.

Good compression will do deduplication. That's kind of how it works.

>posting retard-tier conspiracies on Sup Forums
you need to go back

>Conspiracies
Like the idea that a nation of people just randomly got together and said "oh hey let's execute millions of Jews for no reason whatsoever and we'll do it by building prisoner camps with brothels and dentists and swimming pools then secretly, without anybody noticing during, hoard them into gas chambers". And then for some reason, the leaders (Churchill, Eisenhower, de Gaulle) of the nations that led the fight against these evil, awful, terrible people never mentioned any holocaust or gas chambers just because they forgot to in all their writings.

>Eisenhower never said anything about the holocaust.
Please go back to your retard board.

>Eisenhower talking about bad conditions in camps
Ironic coming from the guy who's forces decided to deny Prisoners of War their rights by just declaring them "disarmed enemy forces". They didn't even build fucking shelters or places to dispose of excrement for their prisoners

>ridiculous claims widely discredited by historians
German PoWs were treated like fucking kings compared to how they treated their prisoners.

Fuck. The fuck off. Back to Sup Forums. You cock dumpsters.

was meant for

magic, dude. Magic.

low entropy data
that is, there's little unique data in the image, so it compresses very well

remember like 10 years ago there were those vidyogammes that were like 300mb to download and too 4 hours to extract and ended up being 8gb on disk?

some repacks do a bunch of tricks to get things smaller, such as;
- recompressing compressed data (such as gzipped textures/map data) with stronger algorithms, to be reversed upon installation
- recompressing video/audio streams and converting them back upon installation

I remember that crazy 56k shooter, shit created textures and models by using basic shapes/patterns and modifying them in a predetermined manner.

you might be thinking of .kkrieger, it's 96k
it uses typical demoscene tricks, procedurally generating most of it's assets at runtime

That's a hashing algorithm though.

Ashing algorithm*

Crackers have been packing games since the 80s, when Rob Nothern hadn't done RNC yet (LZSS+Huffman). We could take games' typically uncompressed program and data files and fit more than one game on a disk, or put two disk games on one disk, or at least usually pack it enough to fit an intro scroller, maybe doc files (someone typed the manual) and a document displayer.

Back in my day you had basic ones like the Happy Computer Packer (Huffman), the excellent Pompey Packer (I think that was an LZ78?), but also better ones like Pack-Ice and (for some stuff, particularly sparse patterned stuff like graphics data) Pack-Fire. All of these were faster than ZIP's deflate when it came around (which hadn't long been put together from LZ77+Huffman by Phil Katz, we had LZH/LZA and ARC before that) but not quite as good overall, but it depends on what you use - you could tweak the tables to pack some things better.

UPX is the only real universal executable packer that survived to this day outside of malware creation, but there are a few specialised ones like Crinkler (packer+linker) for demoscene purposes to help make those small intros/games, and they use every trick they possibly can at the same time (context modelling/mixing, maybe BWT, range/arithmetic encoding) - along of course with lots of procedural generation, up to and including that dank 4k intro shader raycasting magic that's all the rage (and now going out of favour again).

Very cool thanks user

It's an image of a mostly empty disc.

Finally someone who knows some real computer culture here at Sup Forums. Should make a /dsg/, demoscene general one day and see if more people are into it

>windows comes on a dvd
>after install windows takes up like 25GB

So I may be asking something stupid here, but do they take more common letters and attribute smaller binary keycodes to them?

>let's execute millions of Jews for no reason whatsoever
They had their reasons, you don`t kill 6 million for no reason whatsoever.

>half the size for equal quality
HEVC never fails to surprise me doing this

Exactly. If the letters were equally likely to appear this wouldn't do much good, but because some are so much more common than others it's worth the effort to assign them short codes that can't be mistaken for the start of a longer one (look up "prefix code") and let the codes for less common letters get a bit verbose in the process.

Reminds me of that Need for Speed Underground compressed file of ~100mb. I thought it was a ripped version or a virus, but then got it installed and it was the full game.
Granted, it took a while to uncompress, but when I was young I found it impressive as fuck.