ah yes fpgas that you can program using no computers and no software (fpga software is definitely not proprietary software, especially not to this very day)
There are now, at last count, roughly three separate efforts which have achieved open source tooling for various FPGAs. While that was not available at the time, that wouldn't stop anyone from getting the ball rolling... after all, a lack of free compilers certainly didn't stop richard from starting GNU in the first place, now did it? Or do you think he was compiling emacs with a non-existent copy of GCC?
everybody fucks up massively at least sometimes, especially if you've done as much shit as rms has
As much shit? The dude literally has all of three programs to his name, 90% of which is written by someone other than him because he spends most of his time answering emails. He is literally Yanderedev, except Yanderedev is actually coding.
FPGAs are ludicrously expensive for their transistor count, are usually about an order of magnitude slower than purpose-built chips, and generally consume one or two orders of magnitude more power for a given task than purpose-built chips. FPGAs are not, never have been, and shall never be viable for general computing.
All of which is completely irrelevant since the point was to illustrate that free hardware could have been done back then, not that it would have been practical. Free software wasn't practical back then but it didn't stop them from trying (and, for whatever it's worth, it's still a good thing that they did since competition is a good thing).
Technically you can, if you have open source schematics you can bring them to a fabber who will make the chips for you. It's not a very common procedure so it's a pain finding those fabbers. But sometimes you can get a good price, for example if you were a university you could get custom chips for basically free because they sometimes have empty space on wafers if things don't align right
Yeah, MOSIS was a thing even when I was a kid. I'm not sure I want to know how much a run of 40 or so prototypes on a modern node would run... but it would be an interesting case for someone to try and sell corporate bonds to crowdfund an alternative architecture.
they are absolutely not suitable for full general-purpose computing, which is what they were originally brought up as a solution for. try taking an fpga from any given year and running a decent cpu from the same year on it, with a competitive level of performance and power efficiency. good luck!
"Free software isn't feature complete or as optimized as proprietary software! We should totally give up!" - You
Now, if you're content with creating the equivalent of a 6502 to run the equivalent of Xenix, well... things can be done.
also, do you know how to make compiled netlists for e.g. modern xilinx fpgas without their enormous 100+ gb proprietary development toolchain? because i sure don't!
Yes? Have you not been keeping up with the efforts to reverse-engineer their bitstreams? Because I sure have. It's quite exciting. We're at a point where it's possible to break free from their shitty (and I do mean shitty) proprietary toolkit. No matter how incomplete or broken the free/open equivalent is, it's worth the pain to get away from them.
fpgas might be some very neat prototyping devices but i don't think they are, by any means, the final solution to the hardware question
I don't think anyone here was saying that they were. However, they're definitely a solution to someone's needs. I think some people need to understand that not everyone needs the latest whiz-bang silicon from AMD/Intel + Nvidia's monstrous GPUs. For people with more modest needs, old hardware can suit them just fine... and if that hardware is no longer available, FPGAs can provide a free and open solution to that problem.
They may not have been feature-complete or efficient in the mid-1990s, but Spartan-2 or Spartan-3 FPGAs could have been the start of a free hardware revolution even then. Instead, we had to wait for Minimig to show us that open hardware was possible, for the DE-10 to be co-opted into MiSTer, and a dozen autists to prove that free software versions of FPGA tooling was possible. Now here we are... the glorious future where you can plonk down about $500 and get some boards that give you, by and large, circa-2004 hardware that is completely yours. No speculative execution bugs. No hidden Management Engine you can't turn off. None of that bullshit.... and all of it without doubting Stallman who said it couldn't be done.
