>32bit is superior
>posts 16bit OS
the absolute state of AMD64 haters

>>107245257
Nope. 64bit OS on 64bit hardware with 32bit userspace is the best of both.

>>107245257
what kind of shithead midwit "computer" "scientist" said
>no software will address more than 4GB of ram because...because thats what i decided to okay? stop asking questions

>>107245347
Yes, programs should be 32-bit, to conserve RAM and pointers.

>>107245404
Why would any software need more than 4GB?
The only that I know are web browsers, but they are niggered and if they weren't niggered they would use 512MB with 200 tabs opened.
Also, web browsers run in separate processes now anyway, so no need for 64-bit.

Also there were hacks that allowed for more than 4GB addressing in a single process. We could do that rather than niggered 64-bits.

Tell me why a software would need more than 4GB addressing.

>>107245506
>Why would any software need more than an arbitrary amount chosen by some fucking retard
thats a better question. i dont remember people voting for 4GB as universal memory limit.

>>107245506
>Tell me why a software would need more than 4GB addressing.
Because the kernel is typically allocated half of the address space, so it's more like 2 GB. 3 if you enable an option to give it only 1 GB but that breaks some programs.
And then there's other stuff loaded into that address space, so realistically your heap won't be able to get much more than 1 GB.

As for what would need more than that, I've worked with Publisher project files that hit that limit back when it was 32 bit. You can easily get over a GB of stuff when using many high res images to make a large poster.
Yes the program could shift some in and out from disk but that would make significant lag when zooming out or panning around since it needs all of it to render the whole thing. And why should I cuck my performance limiting it to 1 GB when my system has a lot more? With 64 bit it can load it all into the address space, and if physical RAM runs out, the OS can handle paging some out.

>>107245540
>thats a better question. i dont remember people voting for 4GB as universal memory limit.
Funny that we settled for 32bit for so long then.

>you can only use 4GB with 32-bit
holy /g/

>>107245540
The reason why 4GB per process, because existing computers had such architecture and there were no reasons to change it, as you don't need more than 4GB for any process.

>>107245661
>And then there's other stuff loaded into that address space, so realistically your heap won't be able to get much more than 1 GB.
I remember firefox crashing at 1.7GB private bytes. So more than 1GB.

>As for what would need more than that, I've worked with Publisher project files that hit that limit back when it was 32 bit. You can easily get over a GB of stuff when using many high res images to make a large poster.
You used too high res images, you had no reason to use such. Also, the software could implement some kind of compression to conserve the memory.
>Yes the program could shift some in and out from disk but that would make significant lag when zooming out or panning around since it needs all of it to render the whole thing.
Not anymore when we have SSDs.
> And why should I cuck my performance limiting it to 1 GB when my system has a lot more?
You can have unlimited performance at 1GB, the reason why your software used more is because it was poorly programmed.

Also, there were tricks that allowed to use more than 4GB addressing, we should have used them rather than going 64-bit.

>>107245736
Nobody said that.
We are talking about per process addressing limit.
There were hacks and techniques to circumvent the limit, though.

>>107245727
>we settled
who is we?
a dozen shithead boomer programmers who retired long time ago and live as millionaires in some gated community in Santa Clara?

>>107245743
>as you don't need more than 4GB for any process
according to who or what?
who is this "we" and what is this arbitrary 4GB religion morons keep invoking in this thread?

>>107245798
>according to who or what?
Me. I am computer expert and used many kinds of software. No need for more than 4GB per process.
> and what is this arbitrary 4GB religion morons keep invoking in this thread?
When you have 32-bit pointers they can address 4GB of memory. So a single process with a single address space can address 4GB.

64-bit increased pointers to 64-bit, and because of that pointer heavy apps such as web browser are now using 25% more memory for same task.

>>107245763
>who is we?
Mankind, for years.

old = good
modern = bad and gay

>>107245991
Based.

>>107245831
>same task
why do you lie like this?
browsers have 50x more features and options now

>>107245743
>1.7GB private bytes
This includes data sections and heaps allocated by anything else in the address space too. I'm surprised it got that high, but it varied by program.
>You used too high res images
More like I could have cut out areas covered by other elements, but then I'm fucked if I want to move one of those elements.
>Also, the software could implement some kind of compression to conserve the memory
Some image data compresses better than others. Also why trade processor power that's going to hurt screen redraw speed for lower memory usage when I have tons more memory free, if only the program could address it?
>Not anymore when we have SSDs.
Yes anymore with SSDs, they're fast but they're not RAM levels of fast.

>Also, there were tricks that allowed to use more than 4GB addressing, we should have used them rather than going 64-bit.
You mean PAE? Isn't that just for using more total system memory and not per application? If you mean segment-offset addressing or bank switching like was used in the 16 bit era to use more than 64 KB, I've programmed for that shit and don't want to touch it again, it is an abomination. Flat addressing is far superior.

>>107246140
>Yes anymore with SSDs, they're fast but they're not RAM levels of fast.
PCIe 5.0 NVMe SSDs are almost DDR4 level of speed.

>>107245257
>>107245347
This. x32 ABI is based.
https://en.wikipedia.org/wiki/X32_ABI

>>107245988
64bits has been a thing for years too

>>107246511
Yeah, decades before it became norm on consumer market. Yet the 90's were basically all 32-bit for the consumer market, even half of 00's were for most people.

>>107246073
>why do you lie like this?
It's not a lie. On the day when firefox switched from 32-bit to 64-bit, it uses 25% more memory because of bigger pointers.
Same with other programs.
Pointers are double in size.
>browsers have 50x more features and options now
I DIDN'T ASK FOR THE FEATURES. I want HTML4, png/jpeg/gif images, hyperlinks.

>>107246140
>Some image data compresses better than others.
In quality? Yes.
In size? Not always. There are compression algorithms that produce always the same size and compression ratio. For example S3TC texture compression. It has a fixed size per pixel no matter what image you have.
> Also why trade processor power that's going to hurt screen redraw speed for lower memory usage when I have tons more memory free, if only the program could address it?
You shouldn't have a lot of memory, it should be 100$ per GB and limited for every person.
>Yes anymore with SSDs, they're fast but they're not RAM levels of fast.
They don't need to, they are quick enough for swapping.
>You mean PAE? Isn't that just for using more total system memory and not per application?
No, I don't mean PAE. It's per system, not app.
I heard about LARGE ADDRESS AWARE something.
Also, application could run in two or more processes, just like web browsers now.
Yes, it does complicate things. But only few applications would need that. Most don't need more than 1GB.

>>107246547
>the 90' and half of 2000s
so 15 years, meanwhule 64bits have been the standard in consumer land for also 15 years. So?

>>107245257
Man, the raging stupidity is off the charts at this time of day. Is it morning in St. Petersburg already?

>>107246604
So yeah, half of the time we could have been on 64bit we were still on 32bit.

>>107246663
>goalpost move
predictable

>>107246616
Hey Jew, I know you are schizophrenic and mentally ill and unable to do basic math of adding +8 east coast/+9 midcoast/+10 west coast to whatever American timezone you are at because you are so fearful of the cross due to its linking with Christianity. Holy shit Jews are paranoid psychotic bitches fr fr lmao. Its funny how easily they out themselves.

>>107246690
What? Wasn't your point that 32 sucks? So why were WE on 32bit so long?

>>107246826
>conflates present with past
new form of schizophrenia unlocked

>>107246859
So why aren't we coming up with 128bit or hell 256bit computers already to make a leap so we can fully utilize advanced technology like AI in a more efficient way?
>>107246596
Firefox is bloatware with spyware on it 100%.

>>107246859
What's your point then? You haven't posted an actual argument for two posts now.

>>107246944
>So why aren't we coming up with 128bit or hell 256bit computers already to make a leap so we can fully utilize advanced technology like AI in a more efficient way?
We already have those, you probably post from a machine with a CPU that has 128bit instructions.

>>107245257
quack

>>107246971
Nah these are just 64bit Wincucks machines I've been posting from.

>>107246944
>So why aren't we coming up with 128bit or hell 256bit computers already to make a leap so we can fully utilize advanced technology like AI in a more efficient way?
do you have a toddler's understanding of how computers work?
>>107246971
>You haven't posted an actual argument for two posts now.
why should I bother posting an argument when no arguments were presented in the first place?
>What's your point then?
I know it may come as a shock, but.... something that was fine 20 years ago.... might not be good eoight today *mindblown emoji*

File: updooter.png (338 KB, 1398x683)

338 KB PNG

>>107246616
I am posting your photo, jew.

>>107246944
>Firefox is bloatware with spyware on it 100%.
Yes. But it doesn't matter, what matters is that when they switched to 64-bit it instantly increased memory usage by 25%.

>>107247097
>I know it may come as a shock, but.... something that was fine 20 years ago.... might not be good eoight today *mindblown emoji*
It is good, old software and operating systems are better than modern. Even video games, old are better.
OLD = GOOD
new = gay, faggy

WE ARE REGRESSING
COLLAPSE

>>107247185
Oh well I guess you're right, I guess my submarine antenna simulator I need to do use to do actual work for humanity doesn't really need 24GB of RAM!

>>107247234
I don't know about projects like that in big institutions, if they need a lot of RAM they should use special computers with 64-bit pointers.
But in standard programs for real humans at home, I never saw the need for more than 4GB per process.

32-bit are optimal

>>107248346
yes

>>107247700
>people do not need to game at homr

>>107248804
Most games until recently didn't even use 4GB of system memory.
Anyways, if I remember correctly, the 4GB limit is shared with VRAM too?

>>107247185
I'm sure the memory usage wasn't linked in spyware chirping back to homebase faggot lol. Spyware uses up your Memory.

>>107248804
HoMM3 will run just fine

>>107248804
Games should not use more than 1GB.

>>107248846
>Anyways, if I remember correctly, the 4GB limit is shared with VRAM too?
I don't think so.

>>107248908
Not just HoMM3, even big 3D games like GTA IV could run on 32-bit.

>>107249164
GTA V did run on PS3 which has 256MB RAM.

>>107248846
>>Anyways, if I remember correctly, the 4GB limit is shared with VRAM too?
Yeah

>The total amount of RAM and VRAM that the processor can use must fit within this 4GB limit. For example, if a system has 1GB of RAM, it can use an additional 3GB of VRAM, but it cannot use more than 4GB of total memory.

>>107249385
Give source for that shit.

But who needs so much VRAM? GTA IV and V ran on PS3 which has 256MB RAM + 256MB VRAM.

>>107249385
That's with PCIe address space, if you don't use ReBAR it's going to be only like 256MB.

>>107245506
every single thing i do requires more than 4gb of ram. recording my burps and splicing them together with farts for example. cum

>>107249719
Because you are a faggot and use faggy software like Chrome/Firefox/Photoshop/Blender.
I don't do anything that requires 4GB and I am computer expert.

yes

>>107247185
Of all the embarrassing cries for attention and validation. Holy shit.

File: anti-semites.jpg (83 KB, 1680x945)

83 KB JPG

>>107247134
Careful, Marina.

>>107249805
how am I going to load 10 4k female skin textures at once

>>107245257
fuken lmao
most retarded shit I've seen this week

>>107245257
Imagine the inefficiency and power consumption of this old stinky shitbox.

>>107251777
Imagine the input lag, bloat, spyware, of modern computers.

>>107246596
>In quality? Yes.
>In size? Not always. There are compression algorithms that produce always the same size and compression ratio. For example S3TC texture compression. It has a fixed size per pixel no matter what image you have.
This is great for games but why would I ever use lossy compression for materials in high quality print media?
>You shouldn't have a lot of memory
Why? Once you have the required miniaturization, DRAM chips are cheap to produce and more memory just means putting more of them. Meanwhile processors are running up against several barriers put up by the laws of physics.
>LARGE ADDRESS AWARE
This is the thing that pushes the kernel space down to 1 GB, giving the application 3 GB which still isn't all that much if it has a large data set. If the OS is 64 bit this can give an application 4 GB which is a bit better but still not breaking the 32 bit imposed limit.

>>107253892
>but why would I ever use lossy compression for materials in high quality print media?
To save RAM.
>Why?
Because when you have computer with slow CPU and little RAM, only experts and white people are able to make programs for them, this results in high quality programs and operating systems, like in the past.
With a lot of RAM, now niggers and pajeets can do their "apps" in meme languages and technologies like Electron. The results are horrible.
We were better off with the old and slow computers. Even video games were better. No progress but regress.
>This is the thing that pushes the kernel space down to 1 GB, giving the application 3 GB which still isn't all that much if it has a large data set. If the OS is 64 bit this can give an application 4 GB which is a bit better but still not breaking the 32 bit imposed limit.
We could invent some other tricks. One trick is to run two or more processes for the app and communicate between them. Some less important data could be moved to secondary process.

32-bit best.

File: malloc_primer.png (480 KB, 1920x8100)

480 KB PNG

>>107245506
Sometimes it's nice to mmap files bigger than a gigabyte, and it's also nice to reserve large amounts of address space for arenas to prevent heap fragmentation.
ASLR is also basically useless on 32-bit because there's not enough space to randomize addresses over.

>>107245257

if you can write 48-bit depth pictures with 32-bit os would 16-bit or 64-bit os perform faster

>>107245257
Besides the 64-bit architecture, which is garbage, we have the problem of SSDs that fail without warning. I just lost my operating system because my SSD died. Now I am installing a proper HDD hard drive.

>>107245257
i just wish there was a standard language that was cross compatible with all architectures.
not one for 16bit, one for 32, one for 64.
just one, is that so much to ask
>just try [new and exciting new language]
NOOOO

>>107255546
64 bit does a lot of things a bit faster since it has twice as many general purpose registers so variables get stored back to RAM mid computation less often. But that doesn't have anything to do with 48 bpp images, which just means each RGB component is 16 bits instead of 8. And for most operations each component of each pixel is processed separately (or possibly all loaded into SSE together with other pixels if the same math is going to be done on all of them).

>32bit bad, 64bit good
>ipv4 good, ipv6 bad
?

>>107257103
32bit good, ipv4 good

>>107256576
You can restore your OS from backup.

>>107256576
Never had a SSD fail entirely, they just go read-only mode for me.

Also >>107258771

31bit > else
32bit is for plebs.

>>107258806
Wrong, 32 bit can be split in half into 16bit, then 8 bit, then 4bit, then 2 bit, then 1bit. You can pack multiple values into 32bit.
With 31bit you can't do that.

>>107258867
>He think address space for memory needs even division
lol.

>>107258888
It does.
32-bit allows any smaller values to be packed into 32-bit while still having top performance.

>>107246376
i wanna do an x32, -Os install of gentoo just to see how small it can go

>>107246376
Why isn't this default in loonix and windows? 95% of programs don't need 2GB+ of RAM.

>>107259263
not him but x32abi is effectively its own architecture. that means that neither 32bit nor 64bit software can use x32abi libraries, so if you wanted to run 32bit or 64bit software, you need to also install libraries in those architectures, which limits its uses
like; i could build most of my system with x32abi, but if i wanted to use say, wine, i would need to have separate libraries built just for that, since no windows software is built for x32abi. it's not a drop-in replacement

>>107259263
I think only MacOS X used this for a while, with the 64bit system and 32bit userspace model.

>>107259289
Yes, but why aren't software developers choosing x32abi instead of 64bit when developing their software? 64-bit should be dead or rarely used for high memory software.
Almost everything should be in x32abi, no reasons to use 64-bit.

>>107259307
But they don't anymore?

>>107259594
>but why aren't software developers choosing x32abi instead of 64bit when developing their software?
for the same reason i stated. a program built for x32abi needs x32abi support from the OS and x32abi libraries available for it. it's not something you can just drop-in to an existing 32bit or 64bit system. if i gave you an x32abi executable right now, and you weren't already set up for it, you wouldn't be able to run it

>>107259307
>>107259618
-- also, don't confuse this with 32bit software running on a 64bit kernel, it's not the same thing. x32abi software is running in 64bit mode, just with 32bit pointers, so you get the benefits of everything added in AMD64, but without the additional memory overhead of 64bit pointers

>>107259146
-Ofast: “Let’s increase performance of execution by slowing down the loading of the binary because it’s bigger” aka actually slower…

-Os is the fastest then… Ok, time to recompile everything with -mx32 -Os.

>>107259676
i don't pretend to know all their is about compiler optimisation, but i do know that -Os is basically "-O2, but without anything that increases binary size, and maybe with things that decrease binary size", or in other words, it's about keeping binary size down, this can in turn result in lower memory usage, but the details of which i'm not that familiar with

>>107259618
>>107259629
But x32abi should be the default and operating systems should have all libraries for it, because x32abi is better than 64bit.

>>107259748
They shouldn't use consumer computers but special computers made for them. Or some modified/hacked OS.
Also I am sure their software uses tens GB because it's poorly designed and written.

>>107259786
>But x32abi should be the default and operating systems should have all libraries for it, because x32abi is better than 64bit.
maybe it'd have been a logical step from 32bit, but it came after amd64 so it's not that easy. you can of course use it yourself if you want. gentoo supports it

>>107259857
>They don't, and they don't use 32 bit shit either
>Most HPC clusters are using x86_64/ARM Linux
So do you agree that home computers could have stay on 32-bit?
>And I'm sure you're mouth breathing retard. High RAM consumption comes from highly detailed computational mesh.
They don't need that detailed mesh. GTA IV had detailed mesh and it ran on PS3.

>>107259916
>maybe it'd have been a logical step from 32bit, but it came after amd64 so it's not that easy.
This is a common problem in computers, something shit was made earlier and becomes a monopoly, when later someone designs better shit nobody will use it.
>gentoo supports it
I will not use a meme OS.

>>107259984
>This is a common problem in computers, something shit was made earlier and becomes a monopoly, when later someone designs better shit nobody will use it.
in this case it was a bit of a chicken and egg problem. x32abi is a modified version of amd64, so it can't have come first, unless amd themselves thought of it first, but then that's unlikely since even if they had thought of it, they would necessarily have had to provide full 64bit support at the same time as one of the driving points to 64bit was so those programs that /did/ need more than 4GiB of ram could do so
so it's like; sure, in 2003 when the first amd64 cpus came out, x32abi would have been great, faster performance, more registers, no extra memory overhead... but the people buying them probably wanted them to get past the 4G ram barrier. consumers kept using 32bit software, especially in windows, for another decade or so

basically i think the issue is that the benefit isn't seen as significant enough to outweigh the lack of compatibility.
keep in mind work on x32abi for linux only began around 2011, someone had to actually come up with it and implement before it could be considered an option, by 2011 it was firmly a competitor to regular amd64. i'm sure some places use it, there are some applications where it can provide a significant performance improvement, not just because of memory usage but also cpu cache usage, too, like as long as you're using <4GiB per-process, there's no downside performance-wise
>I will not use a meme OS.
it's not a meme os, it's referenced in a meme, but the os itself is not

>>107260065
>basically i think the issue is that the benefit isn't seen as significant enough to outweigh the lack of compatibility.
Maybe because people can simply compile to 32-bit if they want smaller pointers and it will just work on any OS.
>it's not a meme os, it's referenced in a meme, but the os itself is not
It is, it is not designed for productivity and ease of use.

>>107260254
>Maybe because people can simply compile to 32-bit if they want smaller pointers and it will just work on any OS.
sure, if memory was an issue, you use 32bit, but then you miss out on the benefits of amd64, like running an amd64 in 32bit mode means you're limited to ia32's featureset, so like for example several registers are just locked out entirely as you're in a compatibility mode. regular 32bit mode will never be as fast as x32abi, which is why anything performance-oriented goes for regular 64bit mode instead
>It is, it is not designed for productivity and ease of use.
what makes you think that? also i'm not aware of any other distro with x32abi support

>>107260299
>what makes you think that?
I heard it's a meme where you have to compile everything rather than download binaries.
> also i'm not aware of any other distro with x32abi support
What about Microsoft and Apple distros?

>>107245506
>to conserve RAM and pointers
Conserve for WHAT, exactly? Unused RAM is wasted RAM.

>>107260458
>I heard it's a meme where you have to compile everything rather than download binaries.
that's technically optional. while the default action is to compile software from source, you can choose to use prebuilt binaries like any other distro, this was recently-ish (like 2 years ago) expanded to cover most software so your information might be out of date.
>What about Microsoft and Apple distros?
i'm not aware of any x32abi or similar effort with relation to windows or macos.

>>107260472
FOR OTHER PURPOSES. You save RAM by using 32-bit pointers, then you can use the RAM to store more of the other stuff and programs.

Also smaller pointers = more of your program and data will fit into cpu cache = higher performance.

>>107260478
Does gentoo support old GUI with pixelart icons and graphics?

>>107245404
>what kind of shithead midwit "computer" "scientist" said
>>no software will address more than 4GB of ram because...because thats what i decided to okay? stop asking questions
The same one that blames "tech debt" as the reason for why they can't deliver a sprint on time.
Anytime a shithead dev saves that to me, I push back and say, "why wasn't it architected with scale in mind from the getgo?"
This is what separates the code gluers from the actual engineers. They will 100% stammer and fuck around.

>>107260615
>Does gentoo support old GUI with pixelart icons and graphics?
what do you have in mind?

File: wiki-Microsoft-Windows-3-11-n2.png (240 KB, 736x599)

240 KB PNG

>>107260639
>I push back and say, "why wasn't it architected with scale in mind from the getgo?"
Premature optimization is bad. You should develop something quickly and with bad algorithms, then improve if necessary.

>>107260668
Pixelart icons = icons with high contrast, low resolution, visible pixels, low number of colors, saturated colors = Windows 3, Windows 95, Windows 2000.
Modern icons = blurry, gradients, antialiasing, unsaturated colors, low contrast = loonix, modern MacOS.

You're about 20 years late.

File: microsoft-windows^1995^wi(...).png (23 KB, 639x481)

23 KB PNG

>>107260668
Beautiful and soul. Art.

>>107260798
it's amazing how after all this time that's still the most comfortable looking environment i know, and i can visualise what each of those icons do

>>107260776
gentoo can run any de linux can, so it's not really a question of what gentoo can do specifically. you're probably wanting to look at ice2k
https://youtu.be/gDTvalQjfuc

>>107245257
You're like 20 years late bro.

>>107260841
>you're probably wanting to look at ice2k
Looks good but if I run any 3rd party app it will not look in this style, maybe window bar will look correctly but icons inside program, toolbars, won't be in this style.

>>107260870
We can still make the correct decision and go back to 32 bits.

>>107260992
>Looks good but if I run any 3rd party app it will not look in this style, maybe window bar will look correctly but icons inside program, toolbars, won't be in this style.
to an extent, you're right and wrong. the basic visual style of programs follows your toolkit themes, which are customisable, and icon themes are more customisable on linux than they are on windows, like programs will, unless they expressly specify their own custom icon, use your system icon theme.
they layout and positioning of elements of many modern programs probably won't follow that of win95 programs, but neither will they on modern windows, so that's just not something you can expect from any modern os since it's out of their hands

>>107260992
>>107261135
-- like to give you an example, SimpleScreenRecorder there is /not/ part of ice2k, that is a third party program. tell me that looks out of place. it's following the system toolkit and icon theme, like most programs will

>>107260615
>You save RAM by using 32-bit pointers
but then i need indirect/delta pointers if i want to access beyond 4GB of RAM, and that adds unnecessary complexity

>>107261315
not him but his point is that most programs don't need to use >4GiB of ram, which is true. naturally for programs that do you would just use a native 64bit version since that is the most efficient way to run programs that DO need more than 4GiB of ram

>>107261417
>most programs don't need to use >4GiB of ram
It's the Bill of Rights not the Bill of Needs

>>107261315
If your app is so big and important that it uses more than 4GB then the budget is in millions so you can afford some extra complexity.