I think it’s well in valves wheelhouse after proton to do something similar and revolutionise x86 to ARM translation. But at the moment better chips still need to arrive for that too be good enough for a product to built around. Which is why it’s the first thing i think of when they say they need technology to advance more before they make a new steam deck.
x86 to ARM translation is a fairly different problem than what proton solves, so I don’t think it’s clearly in their wheelhouse. Proton / wine is mostly just an implementation of windows libraries on Linux, but doing efficient x86 emulation on arm is a compiler problem. I would guess that Valve could do it or at least hire people to do it, but it’s a bit of a different skill set. Doing x86 efficiently on ARM (particularly with concurrency) also likely involves some extensions to ARM like Apple does with their chips. I haven’t heard if the snapdragon elite chips have anything for x86 compatibility baked in at all. Frankly, I’m treating the snapdragon elite with a fair degree of scepticism until you can actually buy the thing, but I hope it’s good!
Architecture emulation for current gen games is exceptionally unlikely right now. At a fundamental level, wine/proton doesn’t change the instructions the code describes, rather it translates the input and output. It’s a reimplementation of the same instructions in Windows. For architecture crossing you’d either have to create virtual hardware, which adds tremendous overhead, or recompile the binary. Recompilation is theoretically possible, but for x86_64 to ARM64, for games no less, it’s beyond the realm of mortals. It’s like how some jokes can’t be translated between languages; the structure and vocabulary is just too different.
I wonder if the technology they’re waiting for is a more powerful arm processor?
Highly doubt it, because pretty much all games are compiled for x86, and would require dynamic recompilation, which I’m turn costs performance.
Or… they could perform the recompilation beforehand just like the precompiled shaders. Hmmm… that would make it pretty viable!
I think it’s well in valves wheelhouse after proton to do something similar and revolutionise x86 to ARM translation. But at the moment better chips still need to arrive for that too be good enough for a product to built around. Which is why it’s the first thing i think of when they say they need technology to advance more before they make a new steam deck.
x86 to ARM translation is a fairly different problem than what proton solves, so I don’t think it’s clearly in their wheelhouse. Proton / wine is mostly just an implementation of windows libraries on Linux, but doing efficient x86 emulation on arm is a compiler problem. I would guess that Valve could do it or at least hire people to do it, but it’s a bit of a different skill set. Doing x86 efficiently on ARM (particularly with concurrency) also likely involves some extensions to ARM like Apple does with their chips. I haven’t heard if the snapdragon elite chips have anything for x86 compatibility baked in at all. Frankly, I’m treating the snapdragon elite with a fair degree of scepticism until you can actually buy the thing, but I hope it’s good!
No
Did your uncle at valve spill the beans?
Architecture emulation for current gen games is exceptionally unlikely right now. At a fundamental level, wine/proton doesn’t change the instructions the code describes, rather it translates the input and output. It’s a reimplementation of the same instructions in Windows. For architecture crossing you’d either have to create virtual hardware, which adds tremendous overhead, or recompile the binary. Recompilation is theoretically possible, but for x86_64 to ARM64, for games no less, it’s beyond the realm of mortals. It’s like how some jokes can’t be translated between languages; the structure and vocabulary is just too different.