11–20 of 21 posts
Awful. Fortunately, we won't have to suffer x86 much longer.
AMD and Intel have no reason to leave x86. ARM is certainly crowding into their markets, but this isn't the first time, and there's room for multiple architectures.
Awful. Fortunately, we won't have to suffer x86 much longer.
I've been hearing this since the 90s when hollywood was telling me RISC was going to change everything with machines that look crispy in the dark. People were probably saying it in the 80s and I just didn't hear it. AMD and Intel have no reason to leave x86. ARM is certainly crowding into their markets, but this isn't the first time, and there's room for multiple architectures.
Either way there is definitely room for multiple architectures for a long while. Even if one believes ARM or RISC-V or whatever will be the end of x86 it would still take a decade to stop seeing new x86 CPU designs come out just from the momentum.
Earlier quoted context omitted.
I've been hearing this since the 90s when hollywood was telling me RISC was going to change everything with machines that look crispy in the dark. People were probably saying it in the 80s and I just didn't hear it. AMD and Intel have no reason to leave x86. ARM is certainly crowding into their markets, but this isn't the first time, and there's room for multiple architectures.
I wonder how different history might have been if Intel had chose to design something more workable instead of Itanium. We're just now ending the x86-64 transition (no more 32 bit Windows, most all software x86 software is 64 bit native nowadays) and maybe that would never have happened, we'd have just went to 64 bit other-thing. Either way there is definitely room for multiple architectures for a long while. Even if…
x86 will probably get at least a similar lifetime beyond the point where it's clearly dead. And then those systems will likely last another ten years. Maybe there will be retro builds for even longer (Intel Quark was more or less an OG pentium, 20 years later)
Awful. Fortunately, we won't have to suffer x86 much longer.
Be carefull for what you wish for. The design is basically an overextension of the 8008 processer, which assumed 64k of memory and manually programmed in assembly. We've expanded that design for 40 years, and the whole software design philosophy below it has changed. Every design will suffer in these conditions. Meanwhile, almost all info about it is in the open. Culture has closed a lot, even including stuff like RI…
Earlier quoted context omitted.
The reason for the undocumented interface is not because Windows needs to go into protected mode itself (which would be a simple operation that later on came to be available through the more documented DPMI-alikes) , but rather because Windows wants to _replace_ EMM386 itself and "therefore" (usual troll MS logic of the era) needs access to its internals.
I mean, enhanced windows wants to be running in protected mode with its kernel in ring 0. If you're running DOS in real mode, no problem, windows will switch. But, if you've loaded Emm386, it already has the machine in protected mode with its kernel in ring 0. GEMMIS allows switching out the kernel, and later switching back.
Earlier quoted context omitted.
Be carefull for what you wish for. The design is basically an overextension of the 8008 processer, which assumed 64k of memory and manually programmed in assembly. We've expanded that design for 40 years, and the whole software design philosophy below it has changed. Every design will suffer in these conditions. Meanwhile, almost all info about it is in the open. Culture has closed a lot, even including stuff like RI…
It's bizarre to me there are people cheering on the potential death of x86. Sure it has warts, but it is an accidentally created mostly-open platform we'll likely never see created again. And ARM's vaulted power efficiency advantages are really overblown, having more to do with different design goals and not needing to care about a diverse platform than inherent advantages of the underlying cpu architecture. I suspec…
Does this work with multiple processors? It seems like you would have trouble if processor 1 calls cpuid but processor 2 does not for a long time, your emulation of proc 2 would need to be precise to a much-higher-than normal degree to be consistent with proc 1 behavior.
In my case (for a custom hypervisor for a sadly cancelled project) it wasn't a problem because the hypervisor quit itself in early guest boot, which is single CPU only.
Earlier quoted context omitted.
It's bizarre to me there are people cheering on the potential death of x86. Sure it has warts, but it is an accidentally created mostly-open platform we'll likely never see created again. And ARM's vaulted power efficiency advantages are really overblown, having more to do with different design goals and not needing to care about a diverse platform than inherent advantages of the underlying cpu architecture. I suspec…
They are too smart to repeat the mistake of the IBM PC again. Even with "open" ISAs, kernels, they will rules-lawyer out the blackbox part out just like with Android phone makers etc.
Why bother with obfuscated nonstandard schemes when it gets you bad PR and increases your own development costs to get a FOSS stack up and running? Manufacturers just lock the bootloader instead. That can't be reverse engineered around.
Earlier quoted context omitted.
They are too smart to repeat the mistake of the IBM PC again. Even with "open" ISAs, kernels, they will rules-lawyer out the blackbox part out just like with Android phone makers etc.
Weird mobile SoCs are increasingly open. There's device tree. Modern Pixel devices use UEFI. Why bother with obfuscated nonstandard schemes when it gets you bad PR and increases your own development costs to get a FOSS stack up and running? Manufacturers just lock the bootloader instead. That can't be reverse engineered around.
edit: It seems I am misinformed about the moving drivers above part. Perhaps I might have confused this from Fuchsia. In any case the main point was that as you yourself gave the example of, they will always find some part of the stack to hide all the real power/real logic into even if the other parts are open.