Hopefully Qualcomm takes the hint and takes this opportunity to develop a high performance RISC V core. Don’t just give the extortionists more money, break free and use an open standard. Instruction sets shouldn’t even require licensing to begin with if APIs aren’t copyrightable. Why is it OK to make your own implentation of any software API (see Oracle vs. Google on the Java API, Wine implementing the Windows API, etc) but not OK to do the same thing with an instruction set (which is just a hardware API). Why is writing an ARM or x86 emulator fine but not making your own chip? Why are FPGA emulator systems legal if instruction sets are protected? It makes no sense.
The other acceptable outcome here is a Qualcomm vs. ARM lawsuit that sets a precedence that instruction sets are not protected. If they want to copyright their own cores and sell the core design fine, but Qualcomm is making their own in house designs here.
Saying an ISA is just a hardware API vastly oversimplifies what an architecture is. There is way more to it than just the instruction set, because you can’t have an instruction set without also defining the numbers and types of registers, the mapping of memory and how the CPU interacts with it, the input/output model for the system, and a bunch of other features like virtual memory, addressing modes etc. Just to give an idea, the ARM reference is 850 pages long.
APIs can be complex too. Look at how much stuff the Win32 API provides from all the kernel calls, defined data structures/types, libraries, etc. I would venture a guess that if you documented the Win32 API including all the needed system libraries to make something like Wine, it would also be 850 pages long. The fact remains that a documented prototype for a software implementation is free to reimplement but a documented prototype for a hardware implementation requires a license. This makes no sense from a fairness perspective. I’m fine with ARM not giving away their fully developed IP cores which are actual implementations of the ARM instruction set, but locking third parties from making their own compatible designs without a license is horribly anticompetitive. I wish standards organizations still had power. Letting corporations own de-facto “standards” is awful for everyone.
takes this opportunity to develop a high performance RISC V core
They might. This would never be open sourced though. Best case scenario is the boost they would provide to the ISA as a whole by having a company as big as Qualcomm backing it.
RISC V is just an open standard set of instructions and their encodings. It is not expected nor required for implementations of RISC V to be open sourced, but if they do make a RISC V chip they don’t have to pay anyone to have that privilege and the chip will be compatible with other RISC V chips because it is an open and standardized instruction set. That’s the point. Qualcomm pays ARM to make their own chip designs that implement the ARM instruction set, they aren’t paying for off the shelf ARM designs like most ARM chip companies do.
If Qualcomm released a FOSS RISC-V IP core that would’ve required spending multiple millions on hardware engineer salaries (no chance in hell), I would:
In the mobile Linux scene, Qualcomm chips are some of the best supported ones. I don’t love everything Qualcomm does, but the Snapdragon 845 makes for a great Linux phone and has open source drivers for most of the stack (little thanks to Qualcomm themselves).
Hopefully Qualcomm takes the hint and takes this opportunity to develop a high performance RISC V core. Don’t just give the extortionists more money, break free and use an open standard. Instruction sets shouldn’t even require licensing to begin with if APIs aren’t copyrightable. Why is it OK to make your own implentation of any software API (see Oracle vs. Google on the Java API, Wine implementing the Windows API, etc) but not OK to do the same thing with an instruction set (which is just a hardware API). Why is writing an ARM or x86 emulator fine but not making your own chip? Why are FPGA emulator systems legal if instruction sets are protected? It makes no sense.
The other acceptable outcome here is a Qualcomm vs. ARM lawsuit that sets a precedence that instruction sets are not protected. If they want to copyright their own cores and sell the core design fine, but Qualcomm is making their own in house designs here.
Saying an ISA is just a hardware API vastly oversimplifies what an architecture is. There is way more to it than just the instruction set, because you can’t have an instruction set without also defining the numbers and types of registers, the mapping of memory and how the CPU interacts with it, the input/output model for the system, and a bunch of other features like virtual memory, addressing modes etc. Just to give an idea, the ARM reference is 850 pages long.
APIs can be complex too. Look at how much stuff the Win32 API provides from all the kernel calls, defined data structures/types, libraries, etc. I would venture a guess that if you documented the Win32 API including all the needed system libraries to make something like Wine, it would also be 850 pages long. The fact remains that a documented prototype for a software implementation is free to reimplement but a documented prototype for a hardware implementation requires a license. This makes no sense from a fairness perspective. I’m fine with ARM not giving away their fully developed IP cores which are actual implementations of the ARM instruction set, but locking third parties from making their own compatible designs without a license is horribly anticompetitive. I wish standards organizations still had power. Letting corporations own de-facto “standards” is awful for everyone.
They might. This would never be open sourced though. Best case scenario is the boost they would provide to the ISA as a whole by having a company as big as Qualcomm backing it.
The RISCV instruction set IS open source. What they’d do to ratfuck it is lock the bootloader or something.
RISC V is just an open standard set of instructions and their encodings. It is not expected nor required for implementations of RISC V to be open sourced, but if they do make a RISC V chip they don’t have to pay anyone to have that privilege and the chip will be compatible with other RISC V chips because it is an open and standardized instruction set. That’s the point. Qualcomm pays ARM to make their own chip designs that implement the ARM instruction set, they aren’t paying for off the shelf ARM designs like most ARM chip companies do.
BUT Imagine if it was open sourced. God, Gods, by the nine, would be heaven.
If Qualcomm released a FOSS RISC-V IP core that would’ve required spending multiple millions on hardware engineer salaries (no chance in hell), I would:
Simping for Qualcomm is definitely not a take i expected
In the mobile Linux scene, Qualcomm chips are some of the best supported ones. I don’t love everything Qualcomm does, but the Snapdragon 845 makes for a great Linux phone and has open source drivers for most of the stack (little thanks to Qualcomm themselves).
Qualcomm is one of the worst monopolists in any industry though. They are widely known to have a stranglehold on all mobile device development
Or maybe they were just trying to pay a lot less money, and then they got caught at their little trick.
By that logic every company would just run on linux. Free to use ≠ free to implement and support.
You have not read the article.
Do you know how much money you have to pay to make a RISC V chip? Even less than that, since it’s free
If it’s that’s easy / cheap then why have they not?
This is a big ol’ game of bluff from both sides. So, according to you, Qualcomm should call their bluff?
They would need a new core design
Wonder how long that’d take, hmm?
Development is never free, especially, if you have to build new knowhow and can not build upon the one they have built at development of ARM chips.