Yes, it is feasible to build one's own x86 implementation.
Yes, it is feasible to build one's own x86 implementation.
x86 serves as an instruction set, specifying the available commands and their functions. The backend can be built in any way you prefer, provided the frontend remains consistent. Could someone develop a custom version of x86 to run standard operating systems and software? Or would we face limitations due to closed-source constraints?
It seems more of a legal issue than a technical one. Do you own the rights to run everything in x86? The earliest versions aren't covered by patents, but later ones are. If you can't legally make it work, who will bear the cost and effort? A compromise might be designing a processor on another architecture while adding optimizations for faster emulation of x86.
There are still some x86 soft cores available, for instance check these links: https://utoronto.scholaris.ca/items/9498...0af18783fe and https://www.stuffedcow.net/files/henry-thesis-phd.pdf. Remember that x86 supports numerous extensions, many of which are recent and essential for running contemporary programs, though they remain under patent protection.
It doesn't grasp this legal obstacle. If you handled everything on your own using only the public documentation meant for compatibility, why is that an issue? It's similar to Microsoft taking legal action against ReactOS for rewriting the Windows API independently.
It's challenging and I'm not very familiar with the details. I think the original x86 design isn't protected by patents, which means anyone could potentially create it. However, that's quite different from today's x86 architecture, which relies on extensive contributions from both AMD and Intel. To develop a modern feature-level x86 CPU, you'd likely need to obtain licenses from either company or discover an alternative approach that avoids existing intellectual property. It seems smarter to focus on building a non-x86 CPU if you're aiming for current capabilities, rather than getting entangled in patent issues. Emulation could offer a compromise that isn't as contentious.
Of course, that's true. Selling it commercially isn't feasible. It wouldn't stand up to competition without a large team supporting it. As long as you create a processor that executes all instructions in the ISA accurately, you'd have an x86-compliant CPU. However, you'd also need more than just the CPU itself—things like memory controllers and video encoders are essential. You might be better positioned as a company focused on RISC-V cores instead. The licensing isn't about the specific IP blocks but about the ISA itself.