Apple silicon optimization

[spleut]

I've read elsewhere on the forums that optimization for apple silicon isn't likely, but I figured it couldn't hurt to throw out a vote for it anyways. It would also open the door to running factorio on a new ipad which would just be mind boggling.

that's all, thanks for coming to my ted talks.

[ptx0]

same. i don't usually care for Mac hardware/software, but if you could even get an M1 Linux build released, that'd be so excellent.

[ssilk]

I thought a bit about it, and my result was: wube is very aware of that, but there is no big market for them, cause
- Factorio is not streamlined enough (it’s not Minecraft)
- the big players sit in front of PCs
- much too less manpower at wube
- especially Mac/iOS specialists (I’m not sure, but this depends also on the payment, which is in Czech lower and there are eventually not so much Mac users/developers there)
- with growing wube would need to change their whole standing, their whole process, but they like to keep nice and small company (which is nothing bad)

So I don’t think it will come soon.

Off topic:

But so much: I think it’s a mistake. They could learn much about professional software-development if they hire a subteam for Mac/iOS-development. Would be quite expensive, but worth the money, because if they could make a M1-version and it showed up, that this would be - let’s say 3-4 times faster than on the fastest PCs, that would be the last reason to by me a MacBook. And for many others too.

And the same is for iPad: a streamlined version of Factorio (I think to touch instead of mouse or I mean no mods would be allowed by the AppStore) it would be really a hit, I’m very sure.

Off topic 2:

Besides that I think that would be the way to go for this kind of games. The genre of this factory games is only limited by computer power. But factorio was developed 10 years ago and the core of the game is not able to use this cpu-architecture in an optimal way. And I think it would be easier to write a new Factorio instead of trying to optimize it for M1.

The end for CPUs like the intel-chip with different byte length commands is now coming in very long steps. It’s a question of (not so much ) time, that nobody will buy PCs anymore (esp. for gaming), because the architecture of the new CPUs has so many advantages.

[BlueTemplar]

Why would you expect it to be faster ?

On the contrary, Factorio has at this point probably been optimized for x86 in a variety ways, so I would expect the first M1 versions to run several times *slower*.

And optimizing for M1 might require significant changes in the code to the point where Wube would have to deal with two codebases ?

P.S.: And yeah, if you want to run Factorio on an iPad (why?), you also need to jump through Apple's hoops with the App Store. Not to mention that your players would have to buy a mouse & keyboard for their iPads. (And forget about a touchscreen version, that wouldn't work with Factorio - you might as well make a brand new game instead.)

[ptx0]

I thought a bit about it, and my result was: wube is very aware of that, but there is no big market for them, cause

yawn, that's what people say about Linux.

[ptx0]

Why would you expect it to be faster ?

On the contrary, Factorio has at this point probably been optimized for x86 in a variety ways, so I would expect the first M1 versions to run several times *slower*.

And optimizing for M1 might require significant changes in the code to the point where Wube would have to deal with two codebases ?

P.S.: And yeah, if you want to run Factorio on an iPad (why?), you also need to jump through Apple's hoops with the App Store. Not to mention that your players would have to buy a mouse & keyboard for their iPads. (And forget about a touchscreen version, that wouldn't work with Factorio - you might as well make a brand new game instead.)

i think you're missing the point though, running it natively will be faster than Rosetta.

[ssilk]

Why would you expect it to be faster ?

Because the Ram-Speed of M1 is incredible faster. And that is the most limiting factor in bigger factories.
And the graphics might be also much faster, if it uses Metall more directly.

On the contrary, Factorio has at this point probably been optimized for x86 in a variety ways, so I would expect the first M1 versions to run several times *slower*.

Uhm, that’s the point, you need eventually to optimize it. which costs much time. Which costs some money. Which returns only, if you make this effort for MacBook AND iPad.

And optimizing for M1 might require significant changes in the code to the point where Wube would have to deal with two codebases ?

Of course, the core and the gui — because of touchscreen — would be quite different — up to a point where not every (or even no) mods will work/is allowed — but I think the general things don’t need much tuning. Eventually the iPad-variant would be incompatible to play, because the big need to streamline the game for this platform. Eventually! But you and me are no experts. Wube knows that all much, much, much better than us.
I will not discuss this part deeper, because that is too speculative.

All I really can say, that IF they do this step, there is this kind of investment, this type of change of the company, this possible split of compatibility (which may split also the community) and a vage — but possible very high — success for iOS (and macOS).

I really could imagine that it might be a great success. But the chance to fail is also relatively high. For example very much people won’t buy it for 35 € on iPad. Much too much for games on iPad. And so on. As said, there is a big need to streamline it.

P.S.: And yeah, if you want to run Factorio on an iPad (why?),

It’s in its core a really great game. It is difficult, but it will work for other platforms, too. It’s an investment, but it could fail.

you also need to jump through Apple's hoops with the App Store

That’s a nuisance but also helps to market it. And to be honest I think they might make trouble with mods. I don’t know exactly.

Not to mention that your players would have to buy a mouse & keyboard for their iPads.

Of course not. The game needs to change, so that I can play it with my fingers while I sit in a train, while I sit in a train.

(And forget about a touchscreen version, that wouldn't work with Factorio - you might as well make a brand new game instead.)

But that’s needed, as I explained. The ROI will not flow, if you don’t change the GUI. And I don’t see any good reason, why this won’t work.

Maybe this is also the reason why wube has no interest in that: because they think, we cannot make this compatible for iPad. But that’s not really needed. IMHO. Multiplayer will cry loud. That’s not the target. An iPad version of Factorio needs to be much shorter, much more streamlined. And because the core gameplay is generally just a very big mod this is not such a big problem as some might think.

It’s quite difficult of course, but not impossible. And as said, in the end it might need some changes that makes it incompatible. But that’s not the first goal of such a development. The first goal is to get back the investment. IMHO, my opinion.
As said, this is highly speculative. And the point I make here is only: if you don’t insist on being 100% compatible, this is possible.

This is a very important message for an investor. Because if you insist on compatibility this will lead me as software engineer to say “not possible, because too much unknown unknowns”. It opens the mind to think out of the box, which I see at your reaction.

We cannot say this, if we insist on compatibility.

But this brings me to a point: what will the community think of such a step? I don’t know.

[bitwiseshiftleft]

I play on Apple Silicon. I see three issues here: compiling for ARM, Apple's graphics stack, and iPad compatibility.

Recompiling for ARM should be relatively easy unless they're using significant third-party closed-source components. They'd want to make sure that it behaves exactly the same as on Intel (for multiplayer), which would require quite a bit of testing, but it seems entirely doable. But this doesn't help much, because Factorio's weakness on Apple Silicon is the graphics (edit: should say "might not" since IDK how much of the graphics overhead is emulation). Apple's M1 processors are competitive with high-end x86-64 laptop parts in terms of UPS, even when running under Rosetta, but they can't easily sustain 4k at 60 FPS.

Rebuilding the graphics for Apple's Metal graphics stack would be more work, but it would fix the main problem on M1. You'd want a seasoned Apple graphics dev to do it though, and it would take more time, so it might not be worth it for the smaller audience that plays on Mac. On the plus side, this wouldn't touch the game logic, so there wouldn't be any multiplayer compatibility concerns. Also sprite graphics are less of a pain to adapt than something properly 3d. It should be possible for the M1 to absolutely crush Factorio's graphics, at least if it has enough memory. The unified memory system should also make texture streaming unnecessary.

Rebuilding the game for iPad, particularly to run without a mouse and keyboard, would be a large task. It wouldn't be worth doing unless Wube's marketing research says that this market is likely to love Factorio.

[BlueTemplar]

Factorio would suck with touchscreen input because WASD 3rd person control + mouse pointing translates very poorly to it.

You would need to remove the character completely, which then has rippling effects on the rest of the game, at which point you might as well make a different game altogether.
EDIT : Which Wube *could* do, and which *could* be a success, but it would probably end up closer to Mindustry than Factorio. (Note that then porting that potential game to mouse & keyboard inputs *might* be worthwhile, but it will still always feel dumbed down compared to the native ones like Factorio.)

(You can try out Factorio on touchscreen yourself - there are many x86 tablets - I'm running Factorio on one right now, and software adding extra buttons on screen exists too.)

[ptx0]

Factorio would suck with touchscreen input because WASD 3rd person control + mouse pointing translates very poorly to it.

You would need to remove the character completely, which then has rippling effects on the rest of the game, at which point you might as well make a different game altogether.
EDIT : Which Wube *could* do, and which *could* be a success, but it would probably end up closer to Mindustry than Factorio. (Note that then porting that potential game to mouse & keyboard inputs *might* be worthwhile, but it will still always feel dumbed down compared to the native ones like Factorio.)

(You can try out Factorio on touchscreen yourself - there are many x86 tablets - I'm running Factorio on one right now, and software adding extra buttons on screen exists too.)

all this dismissal is pretty fucking irritating. why are you doing this? what do you have to lose by iPad support?

[BlueTemplar]

As ssilk already mentioned - wasted effort by Wube of course ?

You'll notice that it's a qualified dismissal - tell me, haven't you been burned yet by all the failures coming from devs attempting touchscreen <-> mouse & keyboard games/ports (and worse from my point of view than just publisher investment failures : gameplay failures), or haven't you been paying attention to them ? Or maybe you do have actual success stories to share ? (There's Minecraft I guess... but AFAIK it's still dumbed down, simpler than Factorio to start with, and it has overall sold a hundred times more copies than Factorio so it has the advantage of scale.)

I guess that Wube could try to dip their toes into that by first making a x86 touchscreen version of Factorio, and/or an Android x86 version (also m&k ?), both featuring only the sandbox... maybe some character-less scenarios - some additional ones would also be nice to have in the x86 m&k version ?

[ssilk]

To the graphics: Factorio uses its own graphics engine (based on Allegro, which already seems to work for apple silicon https://formulae.brew.sh/formula/openal-soft ).

I think it is easy, but I cannot know.

To Touchscreen: as said you cannot convert Factorio to iPad without changing the gameplay. Character moving and building at the same is maeh.

But — as said — why should it be the exactly same game? Yes, it would be sad, that it is incompatible. But if it is a success, it could be delivered/improved afterwards. Much more important would be to deliver something, that is within the high quality standards of wube and so it is good playable.

[ptx0]

As ssilk already mentioned - wasted effort by Wube of course ?

You'll notice that it's a qualified dismissal - tell me, haven't you been burned yet by all the failures coming from devs attempting touchscreen <-> mouse & keyboard games/ports (and worse from my point of view than just publisher investment failures : gameplay failures), or haven't you been paying attention to them ? Or maybe you do have actual success stories to share ? (There's Minecraft I guess... but AFAIK it's still dumbed down, simpler than Factorio to start with, and it has overall sold a hundred times more copies than Factorio so it has the advantage of scale.)

I guess that Wube could try to dip their toes into that by first making a x86 touchscreen version of Factorio, and/or an Android x86 version (also m&k ?), both featuring only the sandbox... maybe some character-less scenarios - some additional ones would also be nice to have in the x86 m&k version ?

you can connect a bluetooth mouse/keyboard to the thing. it's just that you're afraid for them. don't be. they're far more capable than that.

[BlueTemplar]

Yes, that's what I said, but then *you* dismissed *me*.

[ptx0]

Yes, that's what I said, but then *you* dismissed *me*.

yeah you're all over the place in this discussion which is why i asked why you're doing this and what you have to lose or gain

i didn't dismiss your keyboard idea, i said that you're missing the point that native binaries run faster than rosetta

then you came along and continued to shift the goalpost and try and shut the idea down

so you're projecting now, but it's okay. you haven't changed anything about wube's plans and neither have i.

[spleut]

same. i don't usually care for Mac hardware/software, but if you could even get an M1 Linux build released, that'd be so excellent.

I'm in between gaming computers so I play on my mac these days, so I'm biased. However, one sentiment I have that's independent of whether I'm on mac or not is generally I like to see quality games spreading to all platforms. If you could play the 2nd best rated game on steam (factorio according to steamdb) on an iPad, it would seriously up the ante for all tablet gaming moving forward. I think factorio is in a unique position where they have a phenomenal, well-designed game with great devs that isn't incredibly demanding (barring some megabase type scenarios). I think it has great potential to be on more platforms and would be a great example to all other developers and game studios that top-tier gaming isn't limited to PC.

[Ida-Marie]

I'd raise a hand for M1/M2 optimization too, because of several things:

Optimizing for M1/M2 would mean optimization for ARM, which is coming very fast. For example it is not possible currently to play on a Raspberry Pi since there are no Rosetta for Pi. Raspberry Pi 4 is a potent enough machine for running small factories on.

ARM would be the main computer system in the future, it is just a matter of time before Wube needs to optimize for ARM anyways.

Apple Silicon have the unified infrastructure, which means the game has potential to run much faster and smoother. This would make it possible to reach even larger factories before hitting one of the built-in time-constrains of the game.

Mac is not a bad platform to play Factorio on today, but it could be so much better.

[BlueTemplar]

I'd rather see RISC-V support, since ARM and now even x86 (with hardware DRM support present in Intel and Ryzen processors) are dead-ends to be avoided if you care about general purpose computing.

[ptx0]

I'd rather see RISC-V support, since ARM and now even x86 (with hardware DRM support present in Intel and Ryzen processors) are dead-ends to be avoided if you care about general purpose computing.

off-topic right-wing fear-mongering, SecureBoot was to also be the end of Linux and that never happened

[SupplyDepoo]

The

Efficiency_module_3.png (10.87 KiB) Viewed 5704 times

of the M1 is quite nice. A few years ago I got a Windows laptop with the intention of switching from Mac to Windows to get away from Apple's walled garden and anti-repair practices. Long story short, I'm still a Mac user because I couldn't get used to the way Windows does certain things and felt like, even though I'm 30yo, I need a grandchild to explain things to me. Now I use that laptop only for gaming, but recently with the rising energy costs I decided to measure my usage so I bought one of those cheap watt-meter adapters and started using it on my old fridge, desk lamp, and other doodads. By now I'm confident that it's pretty accurate and yesterday I measured my usage while playing Factorio, first on the Windows laptop, then on the M1.

The results:
Windows: 70W
M1: 20W

To be fair, the Windows laptop is a 17-inch and the M1 is a 13-inch display and a lot of the power goes to that (both measured at max brightness, 300 nits and 400 nits respectively). The M1 also struggles quite a bit with graphics at max zoom level. However, even when idle the Windows laptop consumes an astounding 40W (compared to the M1's 8W). I ran both tests with wifi+bluetooth on, and maxed game settings.